Anda di halaman 1dari 259

Numerical modeling of dam-reservoir interaction

seismic response using the


Hybrid FrequencyTime Domain (HFTD) method

Graduation Committee:
Prof. Dr. A.V. Metrikine
Dr. Ir. M.A.N. Hendriks
Dr. Ir. E.M. Lourens
Ir. W.P. Kikstra
Author
:
Student ID :

Luis Fernando Sirumbal Zapata


4179471

Date

June 2013

Numerical modeling of dam-reservoir interaction seismic


response using the Hybrid FrequencyTime Domain (HFTD)
method

Master of Science Thesis


for the acquisition of the degree
Master of Science in Civil Engineering
at Delft University of Technology

by
Luis Fernando Sirumbal Zapata

June 2013

Graduation Committee:
Prof. Dr. A.V. Metrikine (chairman)
Section of Structural Mechanics, Faculty of Civil Engineering and Geosciences, Delft University of Technology

Dr. Ir. M.A.N. Hendriks


Section of Structural Mechanics, Faculty of Civil Engineering and Geosciences, Delft University of Technology

Dr. Ir. E.M. Lourens


Section of Offshore Wind, Faculty of Civil Engineering and Geosciences, Delft University of Technology

Ir. W.P. Kikstra


Senior Software Development Engineer, TNO DIANA

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

PREFACE

This research work constitutes the graduation project of my studies in the Master of Science in Civil
Engineering program at Delft University of Technology, Faculty of Civil Engineering and Geosciences.
During the last two years in TU Delft, I followed the structural mechanics specialization of the
structural engineering track.
This research project about the Hybrid Frequency-Time Domain (HFTD) method for non-linear
structural dynamics was sponsored and developed in cooperation with TNO DIANA Finite Element
Software Company for Civil Engineering.
I am deeply grateful to Wijtze Pieter Kikstra, Senior Software Development Engineer at TNO DIANA,
because of his unconditional commitment with this project. He is my mentor and friend, and without
his participation the results obtained in the project would not have been as good as they are.
I would like to thank Dr. Gerd-Jan Schreppers, CEO at TNO DIANA, for giving me the opportunity of
work in this project. Special thanks to: Jonna Manie, Software Development Engineer, which
collaboration was crucial in the most important stages of the project; and Maziar Partovi, Advanced
Application Engineer, who was always friendly and eager to talk about life. In general, many thanks to
all the TNO DIANA people with whom I worked and spent enjoyable time during the last eight
months.
From TU Delft I would like to thank my supervisor, Prof. Dr. Andrei Metrikine of the Section of
Structural Mechanics, for his valuable guidance during the graduation committee and personal
meetings we had during the last months. I would also like to acknowledge the important participation
of the graduation committee members, Dr. Max Hendriks and Dr. Eliz-Marie Lourens. And finally I
want to express my gratitude to Prof. Lambert Houben who guided me through all the graduation
process.
I would like to dedicate this work to my home country university, National University of Engineering
(UNI), in Lima, Peru. I am particularly grateful to the people of some institutions, like the Civil
Engineering Faculty Research Institute (IIFIC-UNI) and the National University of Engineering
Foundation (Pro-UNI), who believed in me and supported my studies at TU Delft.
Finally, I proudly dedicate my work to my family, parents and brother, and beloved persons in Peru.
Without their love and constant support it would not be possible for me to be here writing these lines.

L. Fernando Sirumbal Z.
Delft, June 2013

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

vi

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

ABSTRACT

This research work studies the fundamentals of the Hybrid Frequency-Time Domain (HFTD) method
for dynamic analysis of structures, applied to the seismic analysis of dam-reservoir interaction
systems.
The theoretical consistency and accuracy of the HFTD method, based on the frequency domain
Fourier analysis, is demonstrated through all the parts of the thesis, but particularly in Chapter 5 for
linear and highly non-linear single degree of freedom (SDOF) models subject to several types of
loading, and in Chapter 7 for a 2D dam-reservoir interaction finite element model subject to an
earthquake ground acceleration loading.
On the other hand, a Finite Element Method (FEM) formulation of dam-reservoir interaction problems
is presented, being demonstrated the frequency-dependence nature of this type of systems. The fluid
compressibility and boundary conditions of the reservoir are analytically identified as the frequency
dependent properties of the system.
In this context, the HFTD method is proposed as an ideal method for the dynamic analysis of
frequency dependent systems with non-linear behavior, being the dam-reservoir interaction problem
taken as a case study to prove this hypothesis.
Therefore, one important part of the research work is the programming and implementation of the
HFTD method in MATLAB for SDOF models, and in DIANA Finite Element Software for multi degree of
freedom (MDOF) general models. In both cases, the results obtained with the HFTD method are
compared with the solutions of analytical or numerical methods in the time domain, like the wellestablished Newmark method.
Through the entire thesis, special attention is given to the study of the following topics:

Fluid-structure interaction between the dam and the reservoir.


Wave differential equation of a fluid fulfilling the following hypothesis: inviscid, compressible,
small amplitude motion.
Influence of fluid compressibility and reservoir boundary conditions assumptions (f.e: bottom
absorption) in the seismic response of the system.
Soil-structure interaction between the dam and the rock foundation.
Modeling of the far field non-reflecting boundary conditions for the underground and reservoir
(radiation effect).
Modeling of earthquake forces using base acceleration load (earthquake accelerogram).
Numerical implementation issues of Discrete Fourier Transform (DFT) and HFTD (segmenting
approach, time steps, period duration, pseudo force convergence).
Fluid finite elements.
Interface finite elements for fluid-structure interaction analysis.

For MDOF systems the HFTD method is compatible with the Mode Superposition method, which
reduces considerable the size of the problem and therefore the computational time of analysis.
However, the Mode Superposition method also showed to have a great influence in the accuracy of
the results, especially when the HFTD method is applied to non-linear MDOF systems.

vii

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Additionally, the influence in the HFTD performance of other important implementation issues, like the
time segmentation approach, the time step size and the Fourier Transform period, is explained and
numerically demonstrated.
One important conclusion of the research work is that the best performance of the HFTD method, in
combination with the Mode Superposition method, is obtained when it is applied to MDOF systems
with frequency dependent properties and linear or mild non-linear behavior.
From the research work it can also be concluded that these frequency dependent properties actually
have an impact in the response of the dam-reservoir interaction system. However, it was not possible
to outline any definitive conclusion about the relative importance of their inclusion in the systems
response.
Further case studies, especially 3D dam-reservoir interaction models, should be analyzed using the
HFTD implementation in DIANA developed in this research work, with the objective of describing more
clearly the relative importance of the frequency dependent properties in the linear and non-linear
response of this type of structures.
Finally, it is concluded that the HFTD method shows promising perspectives for its application in the
academic and professional practice, especially for the seismic analysis of fluid-structure and soilstructure interaction systems, or any other area for which the research of dynamic non-linear behavior
of frequency dependent systems becomes every time more important.

viii

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Table of Contents

TABLE OF CONTENTS

PREFACE....v
ABSTRACT......vii
TABLE OF CONTENTS......ix
LIST OF FIGURES......xiii
LIST OF TABLES..xix
LIST OF SYMBOLS..xxi
1.

2.

3.

INTRODUCTION.1
1.1

Research project background and motivation.1

1.2

Thesis objectives and research questions..2

1.3

Methodology of work and organization of the thesis..3

FREQUENCY DOMAIN ANALYSIS.5


2.1

Complex frequency response...5

2.2

Direct method...7

2.3

Mode superposition method..7

2.4

Fourier analysis...9
2.4.1.

Complex Fourier series for periodic excitation.9

2.4.2.

Fourier integral for non-periodic excitation...9

2.4.3.

Discrete Fourier Transform (DFT)...10

FINITE ELEMENT METHOD (FEM) SEISMIC ANALYSIS


OF DAM-RESERVOIR INTERACTION PROBLEMS.11
3.1

FEM formulation of dynamic fluid-structure interaction (FSI)..11


3.1.1.

Dynamic FSI in dam-reservoir interaction11

3.1.2.

Eulerian pressure formulation of the fluid wave equation..11

3.1.3.

Reservoir boundary conditions..13

3.1.4.

FEM formulation: Weak form of the coupled system.17

3.1.5.

Discrete coupled problem definition: Galerkin method.18

3.1.6.

Solution of the coupled system in the time domain19

3.1.7.

Solution of the coupled system in the frequency domain...20

3.1.8.

Special case for incompressible fluids...22


ix

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3.1.9.
3.2

4.

Free vibration of FSI systems.23

Modified equation of motion for base acceleration loading...24


3.2.1.

Partitioned discrete equation of motion24

3.2.2.

Superstructure absolute equation of motion..24

3.2.3.

Rigid base movement.25

3.2.4.

Superstructure relative equation of motion.25

HYBRID FREQUENCY-TIME DOMAIN (HFTD) METHOD SOLUTION PROCEDURE27


4.1

Description of the HFTD method..27

4.2

Formulation of the HFTD method for the seismic analysis


of a dam-reservoir dynamic interaction system28

4.3

Implementation issues..32

4.4

5.

Table of Contents

4.3.1.

Solution procedure sequence of the HFTD method...32

4.3.2.

HFTD method time segmentation approach..34

4.3.3.

Fourier transform period...34

4.3.4.

Details of the HFTD method solution procedure


using the time segmentation approach.37

4.3.5.

Definition of the appended decaying functions for the


ground acceleration loading and the pseudo force....40

4.3.6.

Convergence criteria41

HFTD method input parameters42


4.4.1.

Number of modes included in the mode superposition method..42

4.4.2.

Range of excitation frequencies for


the frequency domain solution..42

4.4.3.

Time step size....43

4.4.4.

Parameters for the Discrete Fourier Transform (DFT)..43

SINGLE DEGREE OF FREEDOM (SDOF) DYNAMIC SYSTEMS.........45


5.1

Solution procedure of the HFTD method for SDOF non-linear systems..45

5.2

Time domain analysis of SDOF systems..52

5.3

5.2.1.

Analytic solution in the time domain for SDOF linear systems.53

5.2.2.

Solution procedure of the Newmark method


for SDOF non-linear systems..58

HFTD method numerical test examples for SDOF systems63


5.3.1.

Test examples for SDOF linear systems...63

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.4
6.

5.3.2.

Resonance case study: SDOF linear systems subject


to harmonic loading with exponential amplitude.74

5.3.3.

Test examples for SDOF non-linear systems..80

Effect of non-linearity in the response of SDOF systems...91

SIMPLIFIED ONE DIMENSIONAL DYNAMIC MODELS


FOR DAM-RESERVOIR INTERACTION SYSTEMS........95
6.1

Description of the one-dimensional system...95

6.2

Displacement approach: non-reflective boundary element...96

6.3

Pressure approach: Frequency dependent hydrodynamic pressure..97

6.4

7.

Table of Contents

6.3.1.

Case I: Compressible fluid with radiation


boundary of infinite extent..99

6.3.2.

Case II: Compressible fluid with zero


pressure boundary of finite extent...100

6.3.3.

Case III: Incompressible fluid.101

HFTD method numerical examples..103


6.4.1.

Description of the simplified dam-reservoir interaction system....103

6.4.2.

Dams linear behavior: Influence dam-reservoir


interaction frequency dependent properties...104

6.4.3.

Dams cracking non-linearity: Stiffness softening.106

IMPLEMENTATION OF THE HFTD METHOD IN DIANA FOR THE SEISMIC ANALYSIS


OF FOUNDATION-DAM-RESERVOIR INTERACTION PROBLEMS.113
7.1

Definition of foundation-dam-reservoir interaction models in DIANA...113


7.1.1.

Flow finite elements..113

7.1.2.

Fluid-structure interface..115

7.1.3.

Reservoir boundary conditions116

7.1.4.

Foundation modeling117

7.1.5.

Dams concrete cracking non-linear behavior....117

7.2

HFTD implementation in DIANA....118

7.3

Case study: 2D foundation-dam-reservoir interaction model.119


7.3.1.

Case study description....119

7.3.2.

Eigen-analysis..121

7.3.3.

Frequency domain linear elastic analysis..126

7.3.4.

Transient analysis: Linear and non-linear seismic behavior...129

xi

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

8.

Table of Contents

CONCLUSIONS AND RECOMMENDATIONS........145


8.1

Conclusions...............................................................................................................................145

8.2

Recommendations....147

REFERENCES..149
APPENDIX A: FLOWCHART OF THE HFTD METHOD IMPLEMENTATION IN DIANA.151
A.1.

Main HFTD flowchart..151

A.2.

HRESP function flowchart..152

A.3.

DOTISE function flowchart...155

A.4.

TIMSEG function flowchart156

APPENDIX B: MATLAB ROUTINES..159


B.1.

B.2.

B.3.

HFTD method code159


B.1.1.

Main HFTD script...159

B.1.2.

HFTD selected functions....187

Newmark method code..193


B.2.1.

Main Newmark script...193

B.2.2.

Newmark selected functions.208

Internal force models..218


B.3.1.

Stiffness force functions....218

B.3.2.

Damping force functions...228

APPENDIX C: DIANA HFTD COMMAND FILE..233

xii

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Figures

LIST OF FIGURES

Figure 2.1: Rotating force, displacement, velocity and acceleration vectors in the complex plane.........................6
Figure 3.1: Fluid and solid domains, fluid-structure interface and boundary conditions......................................13
Figure 4.1: Flowchart of the HFTD method general solution procedure.............................................................33
Figure 4.2: El Centro N-S earthquake ground acceleration signal.....................................................................35
Figure 4.3: Spectrum of the ground acceleration amplitude for Fourier
period equal to the loading duration (

)...................................................................................36

Figure 4.4: Spectrum of the ground acceleration amplitude for extended


Fourier period including quiet zone (

)..............................................................................36

Figure 4.5: Graphic representation of the kth time segment.............................................................................37


Figure 4.6: Flowchart of the modified HFTD method solution
procedure including the time segmentation approach.....................................................................................39
Figure 4.7: Modified excitation load with appended decaying
function for the HFTD analysis of the kth time segment...................................................................................41
Figure 5.1: Excitation load, segmentation approach, time span of interest and Fourier period....48
Figure 5.2: Modified excitation load with appended decaying
function for the HFTD analysis of the kth time segment..................................................................................48
Figure 5.3: Single degree of freedom damped system.....................................................................................53
Figure 5.4: Discretization of the force the function..........................................................................................56
Figure 5.5: Harmonic loading example. Displacement response for Analytic and HFTD solutions........................64
Figure 5.6: Harmonic loading example. Velocity response for Analytic and HFTD solutions................................65
Figure 5.7: Harmonic loading example. Acceleration response for Analytic and HFTD solutions..........................65
Figure 5.8: Blast loading...............................................................................................................................66
Figure 5.9: Blast loading example. Displacement response for Direct Integration and HFTD solutions.................67
Figure 5.10: Blast loading example. Velocity response for Direct Integration and HFTD solutions.......................67
Figure 5.11: Blast loading example. Acceleration response for Direct Integration and HFTD solutions.................68
Figure 5.12: Impulsive loading......................................................................................................................68
Figure 5.13: Impulsive loading example. Displacement response for Direct Integration and HFTD solutions........70
Figure 5.14: Impulsive loading example. Velocity response for Direct Integration and HFTD solutions................70
Figure 5.15: Impulsive loading example. Acceleration response for Direct Integration and HFTD solutions..........71
Figure 5.16: El Centro N-S earthquake ground acceleration.............................................................................71
xiii

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Figures

Figure 5.17: Ground acceleration loading example. Displacement


response for Direct Integration and HFTD solutions........................................................................................73
Figure 5.18: Ground acceleration loading example. Velocity
response for Direct Integration and HFTD solutions........................................................................................73
Figure 5.19: Ground acceleration loading example. Acceleration
response for Direct Integration and HFTD solutions........................................................................................74
Figure 5.20: Harmonic sinusoidal load with exponential amplitude for the case study A-4..................................75
Figure 5.21: Resonance case study A-1. Displacement response for Direct Integration and HFTD solutions.........76
Figure 5.22: Resonance case study A-2. Displacement response for Direct Integration and HFTD solutions.........76
Figure 5.23: Resonance case study A-3. Displacement response for Direct Integration and HFTD solutions.........77
Figure 5.24: Resonance case study A-4. Displacement response for Direct Integration and HFTD solutions.........77
Figure 5.25: Resonance case study B-1. Displacement response for Direct Integration and HFTD solutions.........78
Figure 5.26: Resonance case study B-2. Displacement response for Direct Integration and HFTD solutions.........79
Figure 5.27: Resonance case study B-3. Displacement response for Direct Integration and HFTD solutions.........79
Figure 5.28: Half cycle sine pulse loading.......................................................................................................80
Figure 5.29: Linear elasto-plastic stiffness non-linearity...................................................................................81
Figure 5.30: Linear elasto-plastic stiffness force non-linearity.
Displacement response for Newmark and HFTD solutions...............................................................................82
Figure 5.31: Linear elasto-plastic stiffness force non-linearity.
Velocity response for Newmark and HFTD solutions........................................................................................82
Figure 5.32: Linear elasto-plastic stiffness force non-linearity.
Acceleration response for Newmark and HFTD solutions.................................................................................83
Figure 5.33: Linear elasto-plastic stiffness force non-linearity. Hysteretic
stiffness force-displacement relation for Newmark and HFTD solutions.............................................................83
Figure 5.34: Blast loading.............................................................................................................................84
Figure 5.35: Cubic elasto-plastic stiffness non-linearity...................................................................................85
Figure 5.36: Cubic elasto-plastic stiffness force non-linearity.
Displacement response for Newmark and HFTD solutions...............................................................................86
Figure 5.37: Cubic elasto-plastic stiffness force non-linearity.
Velocity response for Newmark and HFTD solutions........................................................................................86
Figure 5.38: Cubic elasto-plastic stiffness force non-linearity.
Acceleration response for Newmark and HFTD solutions.................................................................................87

xiv

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Figures

Figure 5.39: Cubic elasto-plastic stiffness force non-linearity. Hysteretic


stiffness force-displacement relation for Newmark and HFTD solutions.............................................................87
Figure 5.40: Cubic elasto-plastic stiffness force and displacement-dependent damping
non-linearity. Displacement response for Newmark and HFTD solutions...........................................................89
Figure 5.41: Cubic elasto-plastic stiffness force and displacement-dependent damping
non-linearity. Velocity response for Newmark and HFTD solutions...................................................................90
Figure 5.42: Cubic elasto-plastic stiffness force and displacement-dependent damping
non-linearity. Acceleration response for Newmark and HFTD solutions.............................................................90
Figure 5.43: Cubic elasto-plastic stiffness force and displacement-dependent damping non-linearity.
Hysteretic stiffness force-displacement relation for Newmark and HFTD solutions.............................................91
Figure 5.44: Cubic elasto-plastic stiffness force and displacement-dependent damping non-linearity.
Hysteretic damping force-velocity relation for Newmark and HFTD solutions.....................................................91
Figure 5.45: Effect of non-linear behavior. Displacement response for HFTD solution........................................92
Figure 5.46: Effect of non-linear behavior. Velocity response for HFTD solution................................................92
Figure 5.47: Effect of non-linear behavior. Acceleration response for HFTD solution..........................................93
Figure 6.1: Scheme of a dam-reservoir interaction system..............................................................................95
Figure 6.2: Simplified one-dimensional model for dam-reservoir interaction systems.........................................95
Figure 6.3: Simplified one-dimensional model. Displacement approach with radiation boundary.........................97
Figure 6.4: Simplified one-dimensional model. Hydrodynamic pressure approach..............................................98
Figure 6.5: Simplified one-dimensional model. Frequency dependent hydrodynamic pressure load.....................98
Figure 6.6: El Centro N-S earthquake ground acceleration.............................................................................103
Figure 6.7: Dam-reservoir interaction SDOF systems. Displacement response for linear-elastic behavior...........105
Figure 6.8: Dam-reservoir interaction SDOF systems. Velocity response for linear-elastic behavior...................105
Figure 6.9: Dam-reservoir interaction SDOF systems. Acceleration response for linear-elastic behavior.............106
Figure 6.10: Dam-reservoir interaction SDOF systems. Base reaction response for linear-elastic behavior.........106
Figure 6.11: Cubic stiffness softening force-displacement
relationship for concrete cracking non-linear behavior...................................................................................107
Figure 6.12: Dry dam case. Effect of non-linearity in the displacement response..........................................108
Figure 6.13: Dry dam case. Effect of non-linearity in the stiffness force response.........................................108
Figure 6.14: Dam-reservoir interaction with infinite extent radiation boundary
(Case I). Effect of non-linearity in the displacement response........................................................................109
Figure 6.15: Dam-reservoir interaction with infinite extent radiation boundary
(Case I). Effect of non-linearity in the stiffness force response.......................................................................109
xv

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Figures

Figure 6.16: Dam-reservoir interaction with finite extent reflective boundary


(Case II). Effect of non-linearity in the displacement response.......................................................................110
Figure 6.17: Dam-reservoir interaction with finite extent reflective boundary
(Case II). Effect of non-linearity in the stiffness force response.....................................................................110
Figure 7.1: DIANA 2D flow finite elements...................................................................................................113
Figure 7.2: DIANA 2D boundary flow finite elements.....................................................................................114
Figure 7.3: DIANA 3D flow finite elements...................................................................................................114
Figure 7.4: DIANA 3D boundary flow finite elements.....................................................................................115
Figure 7.5: DIANA fluid-structure interface elements.....................................................................................115
Figure 7.6: DIANA fluid-structure interface elements variables.......................................................................116
Figure 7.7: DIANA predefined tension softening functions for total strain crack model.....................................118
Figure 7.8: DIANA predefined compression functions for total strain crack model............................................118
Figure 7.9: DIANA model of the 2D foundation-dam-reservoir interaction case study.......................................119
Figure 7.10: El Centro N-S earthquake ground acceleration signal..................................................................121
Figure 7.11: El Centro N-S earthquake ground acceleration spectrum.............................................................121
Figure 7.12: Eigen-analysis results for the empty reservoir or dry dam case...................................................122
Figure 7.13: Eigen-analysis results for the full reservoir or wet dam case.......................................................123
Figure 7.14: First three mode shapes in the X-direction for
empty (eigen-dry) and full (eigen-wet) reservoir cases..................................................................................124
Figure 7.15: Eigen-analysis results for the empty reservoir
or dry dam case including initial loading conditions.......................................................................................125
Figure 7.16: Eigen-analysis results for the full reservoir or wet dam case including initial loading conditions.....126
Figure 7.17: Direct frequency linear elastic analysis cases.............................................................................127
Figure 7.18: X-direction Displacement Amplitude-Frequency plot of the dams crest........................................127
Figure 7.19: Y-direction Displacement Amplitude-Frequency plot of the dams crest........................................128
Figure 7.20: X-direction Velocity Amplitude-Frequency plot of the dams crest................................................129
Figure 7.21: X-direction Acceleration Amplitude-Frequency plot of the dams crest..........................................129
Figure 7.22: Transient linear and non-linear analysis cases............................................................................130
Figure 7.23: Empty reservoir concrete cracking non-linear response. Dams crest displacement X-direction.......131
Figure 7.24: Empty reservoir concrete cracking non-linear response. Dams crest velocity X-direction...............131
Figure 7.25: Empty reservoir concrete cracking non-linear response. Dams crest acceleration X-direction........132
Figure 7.26: Empty reservoir concrete cracking non-linear response. Cracking pattern at time 4.49 s...............132
Figure 7.27: Incompressible fluid linear elastic response. Dams crest displacement X-direction........................133
xvi

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Figures

Figure 7.28: Incompressible fluid linear elastic response. Dams crest velocity X-direction................................133
Figure 7.29: Incompressible fluid linear elastic response. Dams crest acceleration X-direction.........................134
Figure 7.30: Incompressible fluid linear elastic response. Hydrodynamic pressures at different time steps........135
Figure 7.31: Incompressible fluid concrete cracking non-linear
response. Dams crest displacement X-direction............................................................................................136
Figure 7.32: Incompressible fluid concrete cracking non-linear response. Dams crest velocity X-direction.........136
Figure 7.33: Incompressible fluid concrete cracking non-linear
response. Dams crest acceleration X-direction.............................................................................................137
Figure 7.34: Incompressible fluid reservoir non-linear response.
Dams concrete cracking pattern at different time steps................................................................................138
Figure 7.35: Incompressible fluid reservoir non-linear response.
Displacement and principal strain fields at time step 2.40 s...........................................................................139
Figure 7.36: Compressible fluid with radiation boundary. HFTD non-linear
solution for the dams crest displacement in X-direction................................................................................140
Figure 7.37: Compressible fluid with radiation boundary. HFTD non-linear
solution for the dams crest velocity in X-direction.........................................................................................140
Figure 7.38: Compressible fluid with radiation boundary. HFTD non-linear
solution for the dams crest acceleration in X-direction..................................................................................141
Figure 7.39: Influence of the dam-reservoir interaction and fluid compressibility in
the concrete cracking non-linear response. Dams crest displacement X-direction............................................141
Figure 7.40: Influence of the dam-reservoir interaction and fluid compressibility in
the concrete cracking non-linear response. Dams crest velocity X-direction....................................................142
Figure 7.41: Influence of the dam-reservoir interaction and fluid compressibility in
the concrete cracking non-linear response. Dams crest acceleration X-direction.............................................142
Figure 7.42: Influence of the dam-reservoir interaction and fluid compressibility in the
non-linear response. Dams concrete cracking pattern at different time steps.................................................143

xvii

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Figures

xviii

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Tables

LIST OF TABLES

Table 5.1: Summary of the HFTD method for SDOF non-linear systems...........................................................51
Table 5.2: Iterative procedure of the HFTD method for SDOF nonlinear systems..............................................52
Table 5.3. Coefficients for the recurrence formulas of the Direct Integration Method.........................................58
Table 5.4: Summary of the Newmark method for SDOF nonlinear systems.......................................................61
Table 5.5: Summary of the Modified Newton-Raphson iterative procedure.......................................................62
Table 5.6: Numerical data of Example 4.2 (Craig and Kurdila, 2006)................................................................63
Table 5.7: Parameters of the SDOF linear system subject to harmonic loading..................................................63
Table 5.8: Analytic solution of SDOF linear system subject to harmonic loading................................................63
Table 5.9: HFTD parameters for the SDOF linear system subject to harmonic loading.......................................64
Table 5.10: Numerical data of Example 4.2 (Paz and Leigh, 2004)...................................................................66
Table 5.11: Parameters of the SDOF linear system subject to blast loading......................................................66
Table 5.12: Direct Integration method coefficients for
the SDOF linear system subject to blast loading.............................................................................................66
Table 5.13: HFTD parameters for the SDOF linear system subject to blast loading............................................66
Table 5.14: Numerical data of Example 4.7 (Paz and Leigh, 2004)...................................................................69
Table 5.15: Parameters of the SDOF linear system subject to impulsive loading................................................69
Table 5.16: Direct Integration method coefficients for the SDOF linear system subject to impulsive loading........69
Table 5.17: HFTD parameters for the SDOF linear system subject to impulsive loading.....................................69
Table 5.18: Numerical data of Example 4.9 (Paz and Leigh, 2004)...................................................................72
Table 5.19: Parameters of the SDOF linear system subject to earthquake ground acceleration...........................72
Table 5.20: Direct Integration method coefficients for the SDOF
linear system subject to earthquake ground acceleration................................................................................72
Table 5.21: HFTD parameters for the SDOF linear system subject to earthquake ground acceleration................72
Table 5.22: Numerical data of the SDOF system for resonance case studies.....................................................74
Table 5.23: Numerical data of case study A: Excitation frequencies close to resonance.....................................75
Table 5.24: HFTD parameters for case study A...............................................................................................75
Table 5.25: Numerical data of case study B: Resonance excitation frequency with very low damping ratios........78
Table 5.26: HFTD parameters for case study B...............................................................................................78
Table 5.27: Numerical data of Example 5.5 (Chopra, 2007).............................................................................80
Table 5.28: Parameters of the SDOF system with stiffness non-linearity subject to sine pulse loading................80
xix

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Tables

Table 5.29: Newmark method (average acceleration) parameters for the SDOF
system with stiffness non-linearity subject to sine pulse loading......................................................................81
Table 5.30: HFTD parameters for the SDOF system with stiffness non-linearity subject to sine pulse loading......81
Table 5.31: Numerical data of Example 8.15 (Humar, 2002)............................................................................84
Table 5.32: Parameters of the SDOF system with stiffness non-linearity subject to blast loading........................85
Table 5.33: Newmark method (average acceleration) parameters for the
SDOF system with stiffness non-linearity subject to blast loading.....................................................................85
Table 5.34: HFTD parameters for the SDOF system with stiffness nonlinearity subject to blast loading...............85
Table 5.35: Parameters for the definition of the nonlinear damping force.........................................................88
Table 5.36: Newmark method (average acceleration) parameters for the SDOF
system with stiffness and damping nonlinearity subject to blast loading...........................................................88
Table 5.37: HFTD parameters for the SDOF system with stiffness
and damping nonlinearity subject to blast loading..........................................................................................89
Table 6.1: Numerical data of the one-dimensional dam-reservoir system........................................................103
Table 6.2: Values assigned to the cubic coefficient

of Eq. (6.37) for each analysis case...............................107

Table 7.1: Type of DIANA finite elements used for the case study modeling...................................................119
Table 7.2: Material parameters for the concrete dam....................................................................................120
Table 7.3: Material parameters for the foundation soil..................................................................................120
Table 7.4: Material parameters for the reservoir fluid....................................................................................120

xx

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Symbols

LIST OF SYMBOLS

Interaction surface area


Body forces vector
Generalized fluid matrix multiplying the second time derivate of the pressure vector
Generalized structure mass matrix
Damping coefficient of a single degree of freedom (SDOF) dynamic system
Wave or sonic speed of the reservoir fluid
Critic damping coefficient
Frequency dependent damping coefficient of a SDOF system
Viscous coefficient of non-reflective boundary element
Wave or sonic speed of the reservoir bottom materials
Structure damping coefficient of a SDOF dynamic systems
Damping matrix
Frequency dependent damping matrix
Fluid matrix multiplying the first time derivate of the pressure vector
Constant damping matrix
Structure damping matrix

Fluid-structure interaction added damping matrix


Generalized fluid matrix multiplying the first time derivate of the pressure vector
Generalized structure damping matrix
Modulus of elasticity of the reservoir bottom materials
Generalized structure stiffness matrix
Frequency
Non-linear internal force of a SDOF dynamic system
External force appended decay function
Pseudo force appended decay function
Damping force of a SDOF dynamic system
Stiffness force of a SDOF dynamic system
Frequency increment or interval
Pseudo force appended decay function vector

xxi

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Symbols

Ground acceleration appended decay function vector


Base reaction force vector
Superstructure external load vector
Structure external load vector
Structure nonlinear internal force vector
Maximum sampling frequency
Vibration frequency of the highest mode included in the analysis
Nyquist frequency
Effective generalized load vector
Gravity constant
Reservoir height
Rigid body displacement vector
Rigid body displacement matrix
Stiffness of a SDOF dynamic system
Structure stiffness of a SDOF dynamic systems

Newmark time integration method effective stiffness


Stiffness matrix
Fluid matrix multiplying the pressure vector
Constant stiffness matrix
Structure stiffness matrix
Participation factor in the X direction of the ith mode of vibration
Reservoir length
Mass of a SDOF dynamic system
Effective matrix in the X-direction of the ith mode of vibration
Generalized mass of the ith mode of vibration
Structure mass of a SDOF dynamic systems
Mass matrix
Fluid matrix multiplying the second time derivate of the pressure vector
Structure mass matrix

Fluid-structure interaction added mass matrix


Normal direction
Normal vector
xxii

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Symbols

Number of sampling points for the Fourier Transform


Number of sampling points per time segment
Shape function matrix
External load defined in the time domain
Hydrodynamic pressure
Incident hydrodynamic pressure wave
Harmonic loading amplitude of a SDOF dynamic system
Reflected hydrodynamic pressure wave
Modified external load for the time segment k
Complex external load amplitude of a SDOF dynamic system

Complex amplitude of the incident hydrodynamic pressure wave

Complex amplitude of the reflected hydrodynamic pressure wave

Newmark time integration method effective load increment

Complex external load amplitude vector

Nodal pressure vector

Complex magnitude of the external load defined in the frequency domain

Complex magnitude of the hydrodynamic pressure load


Pseudo force of a SDOF dynamic system
Modified pseudo force for the time segment k

Complex pseudo force amplitude of a SDOF dynamic system


Pseudo force vector

Complex pseudo force amplitude vector

Modified pseudo force vector for the time segment k and iteration j
Fluid-structure interaction matrix
Time
Last converged time step
Load signal duration
Time increment or time step size
Traction surface force
Decay function duration
Time segment duration
Displacement of a SDOF system

xxiii

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Symbols

Base displacement of a SDOF system


Initial displacement of a SDOF system
Relative displacement of a SDOF system
Bottom reservoir displacement
Structure displacement

Complex displacement amplitude of a SDOF dynamic system

Complex relative displacement amplitude of a SDOF system

Complex base displacement amplitude of a SDOF system


Base displacement vector
Ground displacement vector
Structure displacement vector relative to the base
Superstructure displacement vector
Structure displacement vector

Complex displacement amplitude vector

Complex ground displacement amplitude vector

Complex displacement amplitude vector relative to the base

Nodal displacement vector

Modified ground acceleration vector for the time segment k


Displacement amplitude of the SDOF steady state response

Complex displacement amplitude of a SDOF dynamic system


Fluid velocity
Initial velocity of a SDOF system
Velocity vector

Fluid velocity vector

Complex generalized displacement of the ith mode of vibration

Complex generalized displacement amplitude vector


Impedance factor

xxiv

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Symbols

Phase angle (lag) of the response with respect to excitation force


Bottom absorption wave reflection coefficient
Bulk modulus
Newmark time integration method parameter
Newmark time integration method parameter
I

Fluid-structure interface domain

Reservoir bottom domain

Reservoir bottom absorption domain

Reservoir infinite extent domain

Reservoir free surface domain


Distributed damping coefficient per volume
Viscous damping ratio
Hysteretic damping ratio
Density
Constant hydrostatic density
Density of the bottom reservoir materials
Density of the structure
Structure stress tensor
Initial time point of segment k
Fourier period
Quiet zone duration
Fundamental period
Shape vector of the ith mode of vibration
Mode shape matrix
Circular frequency of vibration
Free vibration frequency of a SDOF under damped system
Natural vibration frequency of a SDOF system
Excitation circular frequency
Fundamental excitation circular frequency

xxv

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

List of Symbols

xxvi

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

1.

Introduction

1. INTRODUCTION

1.1. Research project background and motivation


Dynamic analysis of structures can be performed in frequency or in time domain. However both
approaches have the following advantages and drawbacks that should be taken into account when
facing fluid-structure interaction problems:

Frequency domain analysis methods are restricted to harmonic loading input. Besides, it is
only possible to obtain the steady state response of a linear elastic system.
Time domain analysis can be applied with any type of loading input. Besides, it is possible to
obtain the transient response of a system including several types of non-linearities.
However, time domain analysis cannot be applied directly in dam-reservoir interaction
problems taking into account fluid compressibility or radiation boundary conditions. These
kinds of systems are frequency dependent, and therefore, their equation of motion in the time
domain is defined in terms of convolution integrals. Due to their complexity, these differential
equations with convolution integrals are not directly solved by the classical time domain
methods.
One indirect way of dealing with frequency dependent properties in time domain analysis is
following a weakly coupled approach between the solid and fluid equations of motion, by
combining both of them in a global iterative procedure. Nevertheless, this also means a large
computational effort.
On the other hand, dam-reservoir interaction problems, including fluid compressibility and
radiation boundary conditions, are naturally solved in the frequency domain by means of a
decoupled solution procedure of the solid and fluid equations of motion. In general, this is the
case for any system with frequency dependent properties.
Furthermore, frequency domain decoupled solution procedure can be applied in combination
with the mode superposition method, attaining a remarkable reduction of the computational
effort.

For these reasons, hybrid methods that combine advantages of both approaches and diminish their
drawbacks are proposed to solve dynamic dam-reservoir interaction problems, including compressible
fluid and radiation boundary conditions.
In this sense, the research work consists of the study and implementation of the Hybrid Frequency
Time Domain (HFTD) method, applied to the numerical modeling of the linear and non-linear response
of concrete dams under seismic loading, including dynamic dam-reservoir interaction, compressible
fluid and radiation boundary conditions.
The thesis includes programming, codification and implementation of the HFTD method in DIANA
finite element software. The transient results obtained with the HFTD method for a realistic model of a
high concrete dam is studied and compared with the results obtained with other time domain analysis
methods available in DIANA.
Through the entire thesis, special attention is given to the study of the following topics:

Fluid-structure interaction between the dam and the reservoir.


Wave differential equation of a fluid fulfilling the following hypothesis: inviscid, compressible,
small amplitude motion.

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

1.

Introduction

Influence of fluid compressibility and reservoir boundary conditions assumptions (f.e: bottom
absorption) in the seismic response of the system.
Soil-structure interaction between the dam and the rock foundation.
Modeling of the far field non-reflecting boundary conditions for the underground and reservoir
(radiation effect).
Modeling of earthquake forces using base acceleration load (earthquake accelerogram).
Numerical implementation issues of Discrete Fourier Transform (DFT) and HFTD (segmenting
approach, time steps, period duration, pseudo force convergence).
Fluid finite elements.
Interface finite elements for fluid-structure interaction analysis.

1.2. Thesis objectives and research questions


The global objective is to study the HFTD method for the seismic analysis of dam-reservoir interaction
problems, determine its performance, and compare the results obtained with the transient response of
time domain analysis methods. In this way, the feasibility of an extended and generalized use of the
HFTD numerical method for dams seismic response analysis in engineering professional practice will
be evaluated.
A second objective is to study the relevance that frequency dependent properties (fluid
compressibility, reservoir boundary conditions) and far field boundary conditions have in the damreservoir seismic response.
A third objective is to study simplified and ideal models of dams (for example, one dimensional single
degree of freedom systems) which make it possible to have an analytical insight of the results
obtained. The results obtained from these simplified models should provide additional information
related with the topics mentioned in the second objective.
A fourth objective is to program, codify and implement the HFTD method for linear and nonlinear
dynamic analysis of fluid-structure interaction problems in DIANA finite element software.
A fifth objective is to use this implementation for the modeling and analysis of a complex real large
scale case study.
Finally, the last objective is to show that HFTD numerical method provides most of the advantages
offered separately by time and frequency domain methods, and therefore it is an ideal approach for
the solution of dam-reservoir interaction seismic problems.
Based on the five objectives previously presented, the following research questions constitute the
guidelines of the thesis:

What is the influence of the frequency dependent properties (fluid compressibility, radiation
boundary conditions) in the seismic response of dam-reservoir interaction systems?
Which are the most important numerical implementation issues the HFTD method?
What is the general performance (accuracy, stability and convergence) of the HFTD method
for the seismic analysis of dam-reservoir interaction problems?

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

1.

Introduction

Which are the most relevant advantages and drawbacks of the HFTD method for seismic
analysis of dam-reservoir interaction systems when compared with time domain analysis
methods?
How feasible is the extended and generalized use of the HFTD numerical method in
engineering practice for the computational modeling and seismic analysis of real high dams?

1.3. Methodology of work and organization of the thesis


First, the research methodology starts with the literature review of the theoretical background with
special attention to the following topics:

Theory of wave propagation dynamics


Frequency domain analysis
Fourier analysis (DFT and FFT)
Fluid-structure interaction (Mixed displacement-pressure Finite Element Method formulation)
Dam-reservoir interaction (boundary conditions)
HFTD method implementation
Time domain analysis
SDOF linear and non-linear dynamic response
Base acceleration dynamic load
Soil-structure interaction
Finite Element modeling of real scale dam-reservoir interaction problems

Second, as a sensitivity exercise, the response of a single degree of freedom model of a dam on rigid
foundation is studied. HFTD and Newmark methods are implemented and applied in these models and
the results obtained are compared and interpreted. In the case of the HFTD method, compressibility
and radiation boundary conditions are taken into account in some cases, whereas in others they are
neglected. The effects of these properties in the seismic response of this simple model are analyzed.
Third, the programming, codification and implementation of the HFTD method in DIANA are carried
out. A real scale two dimensional dam-reservoir-foundation interaction case study is modeled in
DIANA in order to test the HFTD implementation and compare the results obtained with the Newmark
time domain method available in the software.
Finally, conclusions about the influence and importance of compressibility, boundary conditions and
the method of analysis in the seismic response of dam-reservoir interaction problems are elaborated.
The relevance of HFTD method for this type of problems is evaluated.
The case study is a 100 m. height concrete dam, which foundation material is a flexible soil. In
DIANA, 2D (plane strain) models of this case study are elaborated, putting special attention to the
modeling details and results interpretation. In addition to the seismic time history ground acceleration
loading in the horizontal translational direction, hydrostatic water pressure and gravity loads are
considered as initial conditions.
The contents of the thesis are organized in six chapters. Chapter 2 presents a succinct theory
background review of the frequency domain analysis in single degree of freedom (SDOF) and multi
3

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

1.

Introduction

degrees of freedom (MDOF) systems. The frequency domain analysis theory is developed using
complex numbers notation. Topics like mode superposition method, direct method and Fourier
analysis are presented.
Chapter 3 develops the Finite Element Method (FEM) formulation of dynamic fluid-structure interaction
(FSI) applied to MDOF dam-reservoir interaction systems. The discrete fluid-structure coupled
equations of motion are obtained and the influences of the fluid compressibility and the reservoir
boundary conditions are analytically demonstrated. In addition, the modified equation of motion for
base acceleration loading, together with its underlying assumptions and simplifications, are explained.
Chapter 4 develops the HFTD method formulation applied to the MDOF FEM formulation of the damreservoir interaction system presented in Chapter 3. This comprehensive chapter contains the
background theory of the method, and a detailed description of the method procedure,
implementation issues and input parameters.
In Chapter 5 the dynamic response of SDOF systems is studied. For this purpose, the HFTD and
Newmark methods are developed and programmed in MATLAB especially for these SDOF systems.
The MATLAB routines are used to test several types of linear and non-linear examples, and compare
the results obtained with both methods.
Chapter 6 presents the formulation of a simplified one dimensional dam-reservoir interaction system
including considerations like fluid compressibility and radiation boundary condition. The frequency
dependent terms introduced by the hydrodynamic pressure are added to the SDOF formulation
developed in Chapter 5. Again, the modified HFTD method is used to test a particular example of a
simplified dam-reservoir interaction system, and evaluate the influence of compressibility and radiation
boundary in its linear and non-linear response.
Finally, Chapter 7 presents the results of the HFTD method implementation in DIANA finite element
software. A comprehensive study of the seismic response of 2D foundation-dam-reservoir interaction
case study modeled with DIANA is performed. The modeling process, as well as the analysis results, is
presented in detail for linear elastic behavior and dams concrete cracking non-linear behavior. In
addition to the HFTD method, different types of dynamic analysis, like eigen-analysis, frequency
domain analysis (Direct Method) and time domain transient analysis (Newmark method), are also
carried out.
The conclusions and recommendations of the thesis are presented in Chapter 8. Besides, important
information about the HFTD implementation in DIANA, the MATLAB routines developed for SDOF
systems, and the DIANA HFTD command file for the analysis of the 2D foundation-dam-reservoir
interaction case study, is contained in Appendix A, B and C, respectively.

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

2.

Frequency Domain Analysis

2. FREQUENCY DOMAIN ANALYSIS

2.1. Complex frequency response1


The equation of motion of a viscous damped SDOF system subject to a harmonic excitation is
expressed in its complex form by Eq. (2.1). The excitation force, defined in Eq. (2.1) in terms of its
amplitude ( ) and excitation frequency ( ), is a rotating vector force in the complex plane which two
components are real (cosine) and imaginary (sine).

( )

( )

( )

(2.1)

( )

The solution of Eq. (2.1) (particular solution) is denominated the steady state response of the system,
which in combination with the free vibration solution (complementary solution) and the initial
conditions, provides the total response of the system.
The steady state response of the viscous damped SDOF system defined in Eq. (2.1) is also defined in
the complex plane as expressed in Eq. (2.2). Furthermore, the complex steady state response of the
system can be expressed in terms of the excitation frequency
defined in Eq. (2.3).

and the complex amplitude , as

(2.2)

( )

( )

(2.3)

Eq. (2.3) is differentiated with respect to time to obtain the expressions for the velocity and
acceleration expressed in Eqs. (2.4) and (2.5), respectively. Both responses can be expressed in terms
of the displacement.

( )

( )

(2.4)

( )

(2.5)

( )

The magnitude of the displacement ( ) is a complex quantity defined in Eq. (2.6) in terms of the real
amplitude ( ) and the phase angle or phase lag ( ) with respect to the excitation force. Different
expressions for the displacement, velocity and acceleration are obtained in Eqs. (2.7) to (2.9),
respectively, after replacing the complex amplitude given by Eq. (2.6) into Eqs. (2.3) to (2.5).

( )
( )

(2.6)
(

(2.7)
)

(2.8)

Based on Craig and Kurdilla (2006).


5

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

( )

2.

Frequency Domain Analysis

(2.9)

The systems response and excitation force in the complex plane at time t are illustrated in Fig. 2.1,
which presents a graphical representation of the phase angle between the force and the response, the
magnitude of velocity and acceleration with respect to displacement, and the and rad out-of-phase
angles of velocity and acceleration, respectively, with respect to displacement.

Figure 2.1: Rotating force, displacement, velocity and acceleration vectors in the complex
plane.

The substitution of Eqs. (2.3) to (2.5) into Eq. (2.1) results in the expression of the displacements
complex amplitude given by Eq. (2.10). From this equation, the expressions for the phase angle
and the real displacement amplitude are determined in Eqs. (2.11) and (2.12).

(2.10)

(2.11)

( )
( )

(2.12)

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

2.

Frequency Domain Analysis

2.2. Direct method2


The equation of motion of a SDOF in the frequency domain defined in Eq. (2.13) is obtained after
replacing Eqs. (2.3) to (2.5) into Eq. (2.1). The solution of Eq. (2.13) provides the complex
displacement amplitude ( ) of the steady state response for each excitation frequency ( ).

(2.13)

The extension of Eq. (2.13) to MDOF systems is expressed in Eq. (2.14) in terms of the stiffness
matrix ( ), the damping matrix ( ) and the mass matrix ( ). The displacement amplitude and force
amplitude are defined by the complex vectors and , respectively. Usually, when the excitation
force is defined in the frequency domain, the imaginary part of the force amplitude vector is equal to
zero. However, if the excitation force is defined in the time domain, then a direct Fourier Transform is
required to be expressed in the frequency domain (See Section 2.4), and therefore the force
amplitude vector is complex.
(2.14)

One way of solving Eq. (2.14) is by means of the Direct solution method which converts a n x n
complex system of equations into the 2n x 2n system with real coefficients given by Eq. (2.15). This
double-sized system is obtained as a consequence of splitting the displacement and force amplitude
vectors into their real and imaginary components.

]{

(2.15)

The steady state response of the system, given by the real ( ) and imaginary ( ) parts of the
displacement amplitude vector, is obtained from the solution of Eq. (2.15) for each excitation
frequency .

2.3. Mode superposition method3


For linear systems in which , , and
remain constant, the principle of mode superposition is
applicable. This method requires the previous execution of an eigen-analysis from which a number of
important modes of vibration are selected. Then, the solution of the system is approximated to the
superposition of the solutions corresponding to each selected mode of vibration.
Eq. (2.16) presents the mode shape matrix ( ) containing the m vibration mode shape vectors ( )
selected for the execution of the mode superposition method. The complex displacement amplitude

2
3

Based on TNO DIANA (2011).


Based on TNO DIANA (2011).
7

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

2.

Frequency Domain Analysis

vector ( ) is expressed in terms of the complex generalized displacement amplitude ( ) and the mode
shape matrix by means of Eq. (2.17).

(2.16)

(2.17)

Eq. (2.17) is substituted in Eq. (2.14) to obtain the equation of motion in terms of the generalized
system given by Eq. (2.18). The system of equations is reduced to a m x m system (m < n) by
multiplying Eq. (2.18) times the transpose of the mode shape matrix, as expressed in Eq. (2.19) in
terms of the generalized stiffness ( ), mass ( ) and damping ( ) matrices defined in Eqs. (2.20)
to (2.22), respectively.

(2.18)

(2.19)

(2.20)
(2.21)
(2.22)

The generalized displacement vector ( ) is determined from the solution of the reduced equation of
motion defined in Eq. (2.19), which can be solved for each excitation frequency using the direct
method [Eq. (2.15)]. Then is replaced into Eq. (2.17) to obtain .
A more simplified procedure is possible for a special type of damping matrix . Due to the
orthogonality property of the mass and stiffness matrices with respect to the modal shape matrix,
and

are diagonal matrices. On the other hand, the damping matrix

respect to the modal shape matrix and therefore


to

or

is usually not orthogonal with

is not diagonal. However, when

is proportional

the generalized damping matrix is also diagonal and the system defined by Eq. (2.19) is

uncoupled.
For this particular uncoupled system the solution with the mode superposition method for each
excitation frequency is given by Eqs. (2.23) and (2.24), which are derived from Eqs. (2.17) and
(2.19), respectively.

(2.23)

(2.24)

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

2.

Frequency Domain Analysis

2.4. Fourier analysis4

2.4.1. Complex Fourier series for periodic excitation


The periodic loading function ( ) of a SDOF system is expressed in terms of its harmonic
components by means of the complex Fourier series expansion (also denominated Inverse Fourier
Transform) defined in Eq. (2.25).

( )

The fundamental frequency

[ ( )

is related to the longest period (

(2.25)

) of the harmonic components

contained in the loading function ( ) as expressed in Eq. (2.26). The complex magnitude ( ) of
the loading is defined for each harmonic component by the Direct Fourier Transform expressed in Eq.
(2.27).
(2.26)

( )

( )

(2.27)

2.4.2. Fourier integral for non-periodic excitation


If the loading function

( ) is not periodic, it can be expressed in the frequency domain by means of

a Fourier integral. For doing this the fundamental period


is extended to infinity [Eq. (2.28)] and
therefore the fundamental frequency
becomes a differential term [Eq. (2.29)] and the discrete
frequencies become a continuous variable [Eq. (2.30)].
(2.28)
(2.29)
(2.30)

Furthermore, in order to obtain complex loading magnitude ( ) different than zero when applying
the Direct Fourier Transform defined in Eq. (2.27) ( ) in the denominator tends to infinity), the
change of variable expressed in Eq. (2.31) is introduced.

( )

(2.31)

Based on Craig and Kurdilla (2006).


9

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

2.

Frequency Domain Analysis

After replacing Eqs. (2.28) to (2.31) into Eqs. (2.25) and (2.27), and substituting the summatory
operators by integrals, the Direct and Inverse Fourier Transform for non-periodic loading functions are
defined by Eqs. (2.32) and (2.33), respectively.

( )
( )

( )

( )

(2.32)
)

(2.33)

2.4.3. Discrete Fourier Transform (DFT)


The Fourier integral expressions given by Eqs. (2.32) and (2.33) are used in the determination of the
transient response of a system. However these analytical expressions should be numerically
implemented by means of their discrete counterpart, namely, the Discrete Fourier Transform (DFT).
The formulation of the DFT starts with the sampling of the loading function in
points which uniform
separation is defined by the time step size
. Therefore, the fundamental period ( ) and the
frequency interval ( ) are defined in Eqs. (2.34) and (2.35), respectively.
(2.34)
(2.35)

The continuous time ( ) and circular frequency ( ) variables of Eqs. (2.32) and (2.33) have to be
transformed to their equivalents discrete variables for the DFT formulation. In this sense, the discrete
time points ( ) and frequencies ( ) are defined in Eqs. (2.36) and (2.37), respectively. Similarly, the
differential time ( ) and frequency ( ) is expressed in terms of the time step and frequency interval
by Eqs. (2.38) and (2.39), respectively.
(2.36)
(2.37)
(2.38)
(2.39)

Finally, the substitution of Eqs. (2.34) to (2.39) into Eq. (2.32) and (2.33), and the replacement of the
integrals by summatory operators, provide the expression for the Direct and Inverse DFT defined by
Eqs. (2.40) and (2.41), respectively.

( )
(

[ (

[ ( )

(2.40)

(2.41)

10

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

3. FINITE ELEMENT METHOD (FEM) SEISMIC ANALYSIS OF DAMRESERVOIR INTERACTION PROBLEMS

3.1. FEM formulation of dynamic fluid-structure interaction (FSI)

3.1.1. Dynamic FSI in dam-reservoir interaction


Dynamic FSI problems consist of coupled systems which usually describe different physical
phenomena connected one to the other through a common surface called fluid-structure interface.
The coupling of the systems is produced through this interface where unknown interaction forces are
produced.
Each physical domain is described by its corresponding differential equations of motion, which in most
of the cases are expressed in terms of different dependent variables. These equations are coupled
through the boundary conditions imposed in the fluid-structure interface, making the motion of one
domain dependent on the motion of the others.
For the seismic analysis of dam-reservoir interaction problems, although the fluid displacement
remains small, the interaction with the structure is considerable. The base acceleration load is applied
in the rock foundation, and as a consequence, the soil seismic waves propagate and produce the
structural vibration of the dam. The structural motion of the dam generates pressures and wave
propagation in the reservoir. Consequently, the reservoir waves interact with the dam generating
pressure loads on the dams surface. For this reason, none of the two equations of motion (structure
and fluid) can be solved independently of the other.

3.1.2. Eulerian pressure formulation of the fluid wave equation 1


One way to formulate the coupled FEM equations of motion for FSI problems is using a mixed
displacement scalar potential approach, which defines the solid variables in terms of displacement
degrees of freedom (DOF), and the fluid variables in terms of pressure DOF. This is called the Eulerian
pressure formulation.
One of the advantages of this type of formulation is the simple description of the fluid domain using a
single scalar pressure variable. This reduces considerably the number of variables of the system since
only one DOF per node is required to describe the motion of the fluid domain.
Taking into account that for dam-reservoir interaction problems the fluid motion is not substantial but
small, considerable simplifications can be made in its equation of motion formulation. Those
simplifications are a consequence of the following assumptions only valid for this type of problems:

Small displacement amplitudes


Small velocities (convective effects are omitted)
Inviscid (viscous effects are neglected)
Small compressibility (variation of density is small)

Based on Zienkiewicz et al. (2005), Rizos and Karabalis (2000), and Zienkiewicz and Bettes (1978).
11

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

Therefore, for this type of problems, the linearized Navier-Stokes dynamic equation of the fluid in
terms of density (), velocity (v), pressure (p) and constant body force of gravity (b), is defined in the
vectorial Eq. (3.1).
(

(3.1)

Considering that a small variation of density is assumed, in Eq. (3.1) can be taken out of the
derivative operator and be replaced by the constant hydrostatic density ( 0). Furthermore, if the body
gravity force is neglected the dynamic equation of fluid is reduced to Eq. (3.2).
(3.2)

On the other hand, the linearized continuity or mass conservation equation of the fluid based on the
same assumptions is given by the scalar Eq. (3.3).

(3.3)

Density variation is a consequence of elastic deformability and therefore it involves a change in


pressure. Under the previously stated assumptions, and considering that elastic deformability in fluids
depends on the bulk modulus (), Eq. (3.4) holds.
(3.4)

If Eq. (3.4) is replaced into Eq. (3.3) and then it is differentiated with respect to time, the continuity
equation is expressed in terms of pressure by Eq. (3.5).
(3.5)

Eq. (3.2) is differentiated with respect to space and transformed into a scalar equation by applying the
divergence operator. Then it is combined with Eq. (3.5), eliminating the fluid velocity vector variable
(v) and obtaining the classical scalar wave equation given by Eq. (3.6), that governs the fluid domain
motion. The wave speed (c) is defined in terms of the fluid density and bulk modulus, as shown in Eq.
(3.7).

(3.6)
(3.7)

12

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

3.1.3. Reservoir boundary conditions2


The fluid domain which motion is governed by Eq. (3.6) requires some boundary conditions that
complete the system definition. The boundary conditions describe the coupling of the reservoir with
the dam at the interface and the radiation condition of infinite extent in the upstream direction. The
effect of surface waves in the top of the reservoir and the reflection of pressure waves in the bottom
can also be described by the boundary conditions.
In summary, for a reservoir mass of fluid four types of boundary conditions3 are generally defined as
shown in Fig. 3.1.

Figure 3.1: Fluid and solid domains, fluid-structure interface and boundary conditions
(TNO DIANA BV, 2011).

3.1.3.1.

Fluid-structure interface (I)

The expressions describing the boundary conditions should be expressed in terms of pressure (p)
(essential boundary condition) or pressure gradients in the normal direction to the boundary surface
(

) (natural boundary condition).

Taking this into account, the boundary condition for the fluid-structure interface is obtained by
multiplying the transpose of the normal vector to the boundary surface (nT)4 by Eq. (3.2). Then, the
time derivative of the fluid normal velocity ( ) is prescribed in relation with the pressure gradient in
the normal direction (

), as shown in Eqs. (3.8) and (3.9).


(

(3.8)
(3.9)

Based on Zienkiewicz et al. (2005), Fenves and Chopra (1983), Fenves and Chopra (1984) and TNO DIANA BV
(2011).
3
Generally, for 2D models one of these four boundary conditions belongs to each side of the fluid region.
4
The normal vector n points outward of the fluid domain.
13

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

The coupling between the fluid and the structure is achieved through the motion continuity of both
physical domains. In this sense, the prescribed fluid normal velocity ( ) is coupled with the structure
normal displacement ( ) as expressed in Eq. (3.10).
(3.10)

The final expression of the fluid-structure interface boundary condition is given by Eq. (3.11), after
replacing Eq. (3.10) into Eq. (3.9).

3.1.3.2.

on

(3.11)

Free surface waves (s)

It is possible to specify two types of boundary conditions for the free surface of the reservoir. The first
and simplest one is a consequence of neglecting the effect of the surface waves, prescribing a
pressure equal to zero in the horizontal top free surface, as expressed in Eq. (3.12). This is an
essential or Dirichlet type of boundary condition.
on

(3.12)

The second option for the free surface boundary condition is to use Eq. (3.13) as an approximated
expression to determine the pressure caused by the surface gravity waves. In this case,
is the
vertical elevation (Z-direction) of the fluid relative to the mean surface.
(3.13)

Eq. (3.14) is obtained from Eq. (3.9) after defining z as the normal direction (n).

(3.14)

The combination of Eqs. (3.13) and (3.14) allows the elimination of


surface wave condition given by Eq. (3.15)

3.1.3.3.

on

obtaining the linearized

(3.15)

Radiation boundary of infinite extent (e)

Considering only the fluid wave propagation in X-direction, the general solution of the wave equation
given by Eq. (3.6) is expressed in Eq. (3.16).

14

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

(3.16)

On the radiation boundary, which is theoretically located at an infinite distance from the dam in the Xdirection, no incoming waves are entering into the system (only outgoing waves), and therefore Eqs.
(3.17) and (3.18) describe the pressure wave propagation on this boundary.

(3.17)

)
(

(3.18)

Eqs. (3.19) and (3.20) are obtained after differentiate Eq. (3.18) with respect to space (x) and time
(t), respectively.
(3.19)
(3.20)

After eliminating F from Eqs. (3.19) and (3.20), the radiation boundary condition of infinite extent
given by Eq. (3.21) is obtained.
on

3.1.3.4.

(3.21)

Bottom absorption (b)

Similarly to Eq. (3.6), the wave equation that describes the movement of the bottom materials in
terms of the normal displacement to the bottom surface (
) and wave speed (cr) is defined by Eqs.
(3.22) and (3.23).
(3.22)

(3.23)

Taking into account that in the reservoir bottom only outward propagating waves in the normal
direction are considered, and that there is no incoming propagation waves due to the radiation
condition assumed for the thick layer of bottom materials, the general solution of Eq. (3.22) is given
by Eq. (3.24).

(3.24)

15

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

On the other hand, the equilibrium equation is obtained in terms of the fluid pressure (defined positive
for compression) and the normal stresses in the bottom materials surface, as expressed by Eq. (3.25).
(3.25)

Eq. (3.26) is obtained after replacing Eq. (3.24) into Eq. (3.25), whereas its derivative with respect to
time is given by Eq. (3.27).
(3.26)
(3.27)

The interaction and continuity between the fluid and the bottom materials is ensured by the
compatibility of displacements, required to derive Eq. (3.28) from Eq. (3.14). Eq. (3.29) is obtained
after replacing Eq. (3.24) into Eq. (3.28).
(3.28)

(3.29)

After combining Eqs. (3.27) and (3.29), the term F is eliminated to obtain Eq. (3.30). Finally, the
bottom absorption boundary condition defined by Eqs. (3.31) and (3.32) is obtained after eliminating
Er through the combination of Eqs. (3.23) and (3.30).
(3.30)

on

(3.31)
(3.32)

To determine the constant q in terms of the wave reflection coefficient ( ), the harmonic pressure
wave propagating in the normal direction of the bottom surface is defined as the summation of a
outward incident wave (pi) and an inward reflected wave (pr), as expressed by Eqs. (3.33) to (3.35).
(3.33)
(

(3.34)
(3.35)

16

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

After the substitution of Eqs. (3.33) to (3.35) into Eq. (3.31), the relation given by Eq. (3.36) is
obtained for the bottom boundary (n=0). The expression of q in terms of is given by Eq. (3.37),
obtained after the cancellation of the exponential terms of eq. (3.36).

)
(

(3.36)

(3.37)

Eq. (3.37) is replaced into Eq. (3.31) to obtain the bottom absorption boundary condition expressed
by Eq. (3.38) in terms of the wave reflection coefficient ( ).

on

(3.38)

3.1.4. FEM formulation: weak form of the coupled system5


The equation of motion of the solid part of the system (dam and foundation) is given by Eq. (3.39),
whereas the natural or Neumann boundary condition is given by Eq. (3.40)6.

(3.39)

on

(3.40)

The weak form of the solid part of the system is expressed in Eq. (3.41). Besides, Eq. (3.42) is
obtained after applying integration by parts and the Greens theorem over the term containing the
stress tensor (
) in Eq. (3.41).

]
]

(3.41)
(3.42)

The interaction between the fluid and the structure is introduced in the weak form of the solid
equation of motion [Eq. (3.42)] as a Neumann boundary condition in the interface (I), by relating the
surface traction stresses applied in the fluid-structure interface of the solid domain with the
hydrodynamic pressure (p) defined positive in compression. This interaction boundary condition for
the solid domain is expressed in Eq. (3.43).
(3.43)

5
6

Based on Zienkiewicz et al. (2005) and Zienkiewicz and Bettes (1978).


The normal vector nS points outward of the solid domain.
17

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

After replacing Eq. (3.43) into Eq. (3.42) the final expression of the weak form of the solid part of the
system is obtained in Eq. (3.44). This expression is valid for linear and nonlinear analysis. The most
common nonlinearities, like material or geometric nonlinearities, can be introduced in the terms
and .

(3.44)

With the objective to determine the weak form of the fluid part of the system, its wave equation and
boundary conditions previously defined are reproduced below.
(3.6)

on

(3.11)

on

(3.15)

on

(3.21)

on

(3.38)

The weak form of the fluid part of the system is expressed in Eq. (3.45). Eq. (3.46) is obtained after
applying integration by parts and the Greens theorem over the term
of Eq. (3.45).

]
[

)]

]
(3.45)

[ (

)]

]
(3.46)

3.1.5. Discrete coupled problem definition: Galerkin method7


The displacement and pressure fields are discretized using the approximated expressions given by
Eqs. (3.47) and (3.48), respectively. and are the vectors containing the nodal values for each
degree of freedom of the solid and fluid part of the system, respectively, whereas Nu and Np are the
corresponding shape function matrices.

(3.47)

Based on Zienkiewicz et al. (2005) and Zienkiewicz and Bettes (1978).


18

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

(3.48)

After replacing Eqs. (3.47) and (3.48) into Eq. (3.44), the standard Galerkin discretization applied to
the weak form of the solid equation is defined by Eq. (3.49). For the particular case of linear structural
systems, the discrete equation of motion is transformed as shown in Eq. (3.50).

()

( )

( )

(3.49)
(3.50)

The mass (MS), damping (CS), stiffness (KS) and interaction (R) matrices, as well as the external
(
) and internal (
) force vectors appearing in Eqs. (3.49) and (3.50), are directly obtained from
Eqs. (3.44), (3.47) and (3.48). For instance, the expression of the interaction matrix is reproduced
below in Eq. (3.51).
(3.51)

On the other hand, after replacing Eqs. (3.47) and (3.48) into Eq. (3.46), the standard Galerkin
discretization applied to the weak form of the fluid equation is defined by Eq. (3.52). The expressions
defining the matrices MF, CF and KF are given by Eqs. (3.53) to (3.55), respectively.

(3.52)

(3.53)

(3.54)
(3.55)

3.1.6. Solution of the coupled system in the time domain8


The coupled system of equations defined in the general case by Eqs. (3.49) and (3.52) can be solved
in time domain using time stepping procedures. Both equations have to be first discretized in time
separately and then combined, with the objective of eliminating one of the variables ( ) and solve the
resulting equation for the remaining variable ( ). However, the computational effort required to solve
large coupled systems with time stepping procedures is high due to the unsymmetrical characteristics
of the matrices appearing in the combined equation. Furthermore, unconditional stability of the time
integration algorithms is limited even for linear elastic analysis.

Based on Zienkiewicz et al. (2005).


19

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

For this reason, staggered solution procedures are used as a more efficient alternative to solve fluidstructure problems in the time domain. In this type of procedure both equation are solved
independently by means of an artificial uncoupling. This is achieved introducing an assumed or
predicted value of the vector in Eq. (3.49) and solving this equation independently with time
stepping procedures. Consequently, the solution obtained for the vector is then replaced in Eq.
(3.52) and this equation is solved to obtain a new value of . The iterative procedure is repeated until
convergence for both vectors and is achieved.
The advantage of using staggered process is the reduction of the size of the problem and the
computational effort due to the fact that two smaller and uncoupled (instead of one larger and
coupled) systems of equations are solved separately. However, it is important to mention that in most
of the cases staggered solution procedures are only conditionally stable.
In a staggered solution procedure the modified solid domain equation for nonlinear problems is given
by Eq. (3.56) after replacing the predicted nodal pressure vector (

) into Eq. (3.49).

( )

(3.56)

An incremental iterative scheme (f.e. Newton-Raphson method) and a time integration scheme (f.e.
Newmark method) have to be chosen in order to solve Eq. (3.56) and find an approximated solution
for the nodal acceleration vector in the structure (
), which is introduced in Eq. (3.52) to obtain
the modified fluid domain equation for linear behavior given by Eq. (3.57).

(3.57)

Due to linear behavior of the fluid, the solution of Eq. (3.57) only requires the election of a time
integration method, not being required to follow an iterative procedure. A new vector

is

obtained to be introduced again in Eq. (3.56), and the iterative process is continued until convergence
is achieved. Finally, the whole procedure is repeated for the next time step.

3.1.7. Solution of the coupled system in the frequency domain9


A solution in the frequency domain is only valid for linear behavior of the structure. The coupled
system defined by Eqs. (3.50) and (3.52) can be compactly expressed by Eq. (3.58).

]{ }

]{ }

] { }

( )}

(3.58)

It is possible to determine the steady state solution of this coupled system of equations in the
frequency domain by expressing the nodal pressure vector ( ) in terms of the nodal structural

Based on TNO DIANA BV (2011).


20

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

displacement vector ( ). First, the external force vector should be expressed or transformed into the
harmonic form shown in Eq. (3.59).

( )

(3.59)

Considering that both, the structural displacements and the fluid pressures, are produced by the
external force, the steady state solution of both responses will exist in the same harmonic form as
expressed in Eqs. (3.60) and (3.61).

( )

(3.60)

( )

(3.61)

After replacing Eqs. (3.59) to (3.61) into Eq. (3.58), the system of equations shown in Eq. (3.62) is
obtained, where the complex amplitudes of the nodal structural displacements ( ) and pressure ( )
vectors are the unknown variables.

]{ }

(3.62)

From the second row of Eq. (3.62) the amplitude of the nodal pressure vector can be expressed in
terms of the amplitude of the nodal displacement vector as defined in Eqs. (3.63) and (3.64).

(3.63)

( )

(3.64)

After replacing the expression of given by Eq. (3.63) into the first row of Eq. (3.62), the uncoupled
equation to determine the steady state response of the structure in the frequency domain, in terms of
the complex amplitudes of the nodal displacement vector ( ), is given by Eqs. (3.65) to (3.67). The
effect of the reservoir interaction in the dams equation of motion is represented by and , called
added mass and added damping, respectively.

(
(

(3.65)
(3.66)

)
(

(3.67)

Eqs. (3.65) to (3.67) can be solved in the frequency domain using the Direct or the Mode
superposition methods (see Chapter 2).

21

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

3.1.8. Special case for incompressible fluids10


If compressibility and surface gravity waves boundary condition are neglected, a simplified uncoupled
equation describing the fluid-structure interaction is obtained instead of the coupled system of
equations defined by Eqs. (3.49) and (3.52).
Incompressible fluid assumption (
) combined with a prescribed pressure equal to zero in the
horizontal top free surface [Eq. (3.12)], produce null matrices MF and CF, as expressed by Eqs. (3.68)
and (3.69), respectively.
(3.68)
(3.69)

Replacing Eqs. (3.68) and (3.69) into Eq. (3.52) results in the simplified explicit discrete equation
given by Eq. (3.70), which describes the incompressible fluid motion.

(3.70)

In turn, the substitution of Eq. (3.70) into Eq. (3.49) generates a fully uncoupled equation for the
structural part of the system. This simplified structural equation of motion, based on the disregard of
compressible fluid and surface gravity waves, is expressed in Eq. (3.71). It is interesting to note that
the only difference between Eq. (3.71) and the isolated structural system [Eq. (3.49)] is the presence
of the added mass matrix ( ), which expression is given by Eq. (3.72). The simplified Eq. (3.71) for
incompressible fluids can be solved directly in time domain using direct integration or mode
superposition methods, without the necessity of staggered solution procedures.

()

( )

(3.71)
(3.72)

It is not completely clear when it is recommendable to neglect compressibility effects and assume this
simplified special case. One criterion for taking a decision is to estimate the order of magnitude of the
fundamental compressible period of the structure in terms of the dams height (H) and the fluid wave
speed velocity (c), using the expression H/c. If this period is close to the magnitude of the earthquake
acceleration period, then the compressibility effects play an important role and therefore should not
be neglected. This is generally the case for high flexible dams containing a long range of modal
vibration periods. On the other hand, compressibility can be less important in the case of low height
stiffer dams with a compressible fundamental period far smaller than the excitation period.

10

Based on TNO DIANA BV (2011) and Zienkiewicz et al. (2005).


22

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

3.1.9. Free vibration of FSI systems11


If a mode superposition method is used to solve Eqs. (3.65) or (3.71), then first it is required a free
vibration eigen-analysis. Numerous procedures and assumptions can be followed in order to find the
eigen-vectors (or mode shape vectors) and eigen-values (or modal frequencies) of the fluid-structure
interaction system. Nevertheless, several authors, like Fenves and Chopra (1983) and Darbre (1996),
coincide in disregarding the influence of the fluid interaction to determine the mode shapes of the
structure.
It is understood that the shape modes of a structure, like an arch dam, is almost the same no matter
if it is with a full or an empty reservoir. For this reason Eq. (3.50) is used for deriving the free
vibration problem described by Eq. (3.73) and calculating the mode shape vectors of the structure in a
fluid-structure interaction system.

(3.73)

This approach, known as dry dam eigen-analysis, is accurate enough for the calculation of the mode
shape vectors, but not for the calculation of the modal frequencies, which require the consideration of
the fluid effects. One way of doing this is using the full coupled equation defined in Eq. (3.58) to
obtain the free vibration problem expressed in Eq. (3.74).

]{ }

] { }

{ }

(3.74)

However, the matrices of Eq. (3.74) are neither symmetric nor positive definite, which make the
classical computational methods for eigenvalue analysis not directly applicable. Some manipulations
and variable symmetrizing procedures can be followed to solve the eigen-value problem defined by
Eq. (3.74), yet making the process too complicated.
Another approach known as wet dam eigen-analysis consists in assuming that the interacting fluid is
incompressible. Then, the simplified equation of motion given by Eq. (3.71) can be used to determine
the modal frequencies of the fluid-structure interaction system through the solution of Eq. (3.75). The
influence of the fluid is concentrated in the added mass term ( ) given by Eq. (3.72).

11

(3.75)

Based TNO DIANA BV (2011), Zienkiewicz et al. (2005), Fenves and Chopra (1983) and Darbre (1996).
23

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

3.2. Modified equation of motion for base acceleration loading 12

3.2.1. Partitioned discrete equation of motion


In earthquake analysis the input load is defined as a base acceleration which is multiplied by the mass
of the structure, generating an inertial load applied through all the structure. Due to the prescribed
movement applied to the base, the response of the other part of the structure (superstructure) can be
classified as absolute or relative to the base motion.
Generally, earthquake analysis requires the relative displacements, velocities and accelerations for the
design and evaluation of the structure. Therefore, in this section the classical equation of motion for
seismic analysis, in terms of the relative displacement, velocity and acceleration vectors, is derived
and the assumptions relying on this formulation are explained.
When an acceleration load is applied in the base of the structure, the dynamic equation of motion of
this system can be partitioned into known DOF located in the base ( ) (where accelerations are
prescribed) and unknown DOF located in the superstructure ( ). The partitioned equation of motion
is expressed in Eq. (3.76), which right-hand side has no additional external loads in the
superstructure, taking into account that the load is generated by the base motion ( ), and that
is
the associated unknown reaction force vector in the base.

]{

]{

]{

{ }

(3.76)

3.2.2. Superstructure absolute equation of motion


After taking from Eq. (3.76) the first partition belonging to the superstructure, the equation of motion
in terms of the absolute response is defined by Eqs. (3.77) and (3.78).

(3.77)

(3.78)

It can be noticed from Eq. (3.78) that the external force in the absolute equilibrium equation is
produced by the base displacement, velocity and acceleration. However, there are two assumptions
that simplify the force vector given by Eq. (3.78). The first assumption is that in the mass matrix there
is no coupling between the base and superstructure DOF, and therefore Eq. (3.79) holds. The second
assumption, expressed in Eq. (3.80), is to disregard the damping term contribution due to its not
completely clear physical meaning (Wilson, 2002).
(3.79)
(3.80)

12

Based TNO DIANA BV (2011) and Wilson (2002).


24

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

After replacing the assumptions expressed in Eqs. (3.79) and (3.80) into Eq. (3.78), the force vector
of the absolute equilibrium equation is expressed by Eq. (3.81) in terms of base displacement only.
(3.81)

3.2.3. Rigid base movement


If the base of the structure is rigid, then the displacements, velocities and accelerations are uniform
through all the base points in the three directions of analysis. This uniform motion is defined by the
ground displacement vector ( ) as expressed in Eq. (3.82).

(3.82)

The matrices ( ) and ( ) represent the unitary rigid body displacements of the superstructure and
the base, respectively, in the three translational directions (X, Y, Z). Eqs. (3.83) and (3.84) show that
each of these rigid body motion matrices is composed by three vectors (one per translational
direction) which values vary between 0 and 1, depending on the cosine direction of each DOF with
respect to the XYZ coordinate system.
( )

( )

( )

( )

( )

( )

( )

(3.83)

( )

(3.84)

Eq. (3.85) shows the superstructure absolute displacement expressed in terms of the rigid body
motion matrix, the ground displacement and the relative displacement ( ). Similarly, Eq. (3.86)
shows a similar expression for the base, with the obvious difference that in this case there is no
relative displacement.
( )

(3.85)

( )

(3.86)

3.2.4. Superstructure relative equation of motion


The relative equation of motion is obtained by rewriting the absolute equation of the superstructure in
terms of the relative displacements. Therefore, Eqs. (3.85) and (3.86) are replaced into Eqs. (3.77)
and (3.78) to obtain the full relative equilibrium equation of motion given by Eq. (3.87).

( )

( )

( )

( )

( )

( )

(3.87)

25

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

3. Finite Element Method (FEM) Seismic


Analysis of Dam-Reservoir Interaction Problems

Taking into account that internal forces are only proportional to relative displacements (stiffness force)
and velocities (viscous damping force), the multiplication of the stiffness (K) and damping (C)
matrices by the rigid body displacements (I) is equal to zero13, as expressed in Eqs. (3.88) and (3.89).
( )

( )

( )

(3.88)

( )

(3.89)

After inserting Eqs. (3.79), (3.88) and (3.89) into Eq. (3.87), the dynamic equilibrium equation given
by Eq. (3.90) is obtained in terms of the superstructure displacements relative to the base motion
( ). This approximate form of the relative equation of motion is the classic equation used in seismic
analysis.

( )

(3.90)

The base acceleration load can be expressed for each of the three translational directions as shown in
Eq. (3.91), after replacing Eqs. (3.82) and (3.83) into Eq. (3.90).

( )

( )

( )

(3.91)

It is interesting to notice that in the approximate absolute equation of motion [Eqs. (3.77) and
(3.81)], the force vector is proportional to the prescribed base displacements, and it is applied only in
the superstructure nodes close to the base (
). On the other hand, the force vector of the
approximate relative equation of motion [Eq. (3.90)] is proportional to the prescribed base
accelerations, and it is applied in the superstructure nodes not linked with the base (

( )

).

13

For the stiffness matrix this statement is always correct. However, in the case of the damping matrix the statement does not
fulfill if Rayleigh damping definition (aM+bK) is used. The multiplication of the mass proportional term (a0) by the rigid body
motion matrix results in a damping forces different than zero. Nevertheless, this artificial mass proportional damping energy
added to the system is usually ignored (TNO DIANA BV, 2011).

26

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

4. HYBRID FREQUENCY-TIME DOMAIN (HFTD) METHOD SOLUTION


PROCEDURE

4.1. Description of the HFTD method1


The HFTD method is generally applied in the solution of dynamic problems of systems with frequency
dependent properties and non-linear behavior. Frequency dependent properties, which are present in
soil-structure and fluid-structure interaction systems, are properly taken into account by frequency
domain solution methods.
Nevertheless, this type of analysis methods is only applicable in linear systems. When some kinds of
non-linearities (f.e.: material, geometrical, contact, friction, sliding) are required to be included in the
analysis, frequency domain methods are not applicable anymore and a time-domain solution
(generally by means of time-stepping algorithms) should be used.
Time domain analysis methods allow tracking the development of the non-linear response and
evaluating its magnitude in each time step. However, it is not possible to consider frequency
dependent properties directly and consistently.
It is under these circumstances that HFTD method is proposed as an alternative to study the nonlinear dynamic response of frequency-dependent systems, by means of a hybrid formulation that
combines the advantages of both, time and frequency domain analysis methods, and diminish their
corresponding drawbacks.
In general terms, the HFTD method transforms the non-linear equation of motion defined in the time
domain into a pseudo-linear equation of motion defined in the frequency domain. For this to be
achieved, the internal force vector is divided into linear and non-linear components.
The linear component of the internal force is the summation of the linear elastic stiffness and damping
forces. The non-linear component is an unknown vector denominated pseudo force. Besides, the
linear component of the internal force is kept in the left hand side of the equation of motion, whereas
the pseudo force vector is moved to the right hand side. In this way, the left-hand side is composed
exclusively by linear terms, whereas the right-hand side contains the applied external forces and the
pseudo force vector which unknown value is determined iteratively.
As mentioned before, the HFTD method solves iteratively the pseudo-linear (or linearized) equation of
motion in the frequency domain. However, the dynamic external force applied to the system is defined
by a signal in the time domain. In the same way, the pseudo force vector representing the nonlinear
behavior of the system can only be evaluated in the time domain. Therefore, the HFTD method
requires the transformation of both time dependent components of the right-hand side of the
equation (time-dependent loading and pseudo force) into a series of harmonic forces by means of a
Direct Fourier Transform (transformation to the frequency domain).
As a consequence the linearized equation of motion in the frequency domain can be solved for a
range of loading frequencies, using the Direct or Mode superposition methods (see Chapter 2). The
next step is to transform the solution obtained (displacements, velocities and accelerations) back to
the time domain by means of the Inverse Fourier Transform.
1

Based on Darbre (1996) and Fenves and Chavez (1990).


27

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

With the response in the time domain it is possible to use the constitutive laws ruling the non-linear
behavior of the system and evaluate the internal forces. Then, a new value of the pseudo forces is
computed as the difference between the linear elastic force and the nonlinear internal force.
Finally, the new value of the pseudo force vector is transformed again to the frequency domain to
solve one more time the linearized equation in the frequency domain. The process is repeated until
convergence is achieved.

4.2. Formulation of the HFTD method for the seismic analysis of a dam-reservoir
dynamic interaction system2
A general non-linear equation of motion of the solid part of a fluid-structure interaction problem can
be derived from Eq. (3.49). This equation of motion in absolute coordinates is shown in Eq. (4.1).

(4.1)

( )

For the particular case of earthquake analysis of dam-reservoir interaction systems, the seismic load
can be applied as a base acceleration loading in the three translational directions, defined by the
ground acceleration vector ( ). This approach is conveniently formulated in relative coordinates

( )
without necessity of the application of an external force (
). Based on Eqs. (3.90) and
(4.1), the non-linear equation of motion in relative coordinates for the seismic analysis of a damreservoir interaction system is given by Eq. (4.2).

( )

(4.2)

( )

In the HFTD method, the internal force (


) of the dam (structural part of the coupled system) is
decomposed into a linear and a nonlinear components, as expressed in Eq. (4.3). The linear
component is expressed in terms of the initial tangent stiffness and damping matrices of the dam
). On the other hand,
(
), and the displacement and velocity vectors relative to the base (
the nonlinear component is treated as a general remaining force depending on time ( ).

(4.3)

( )

After replacing Eq. (4.3) into Eq. (4.2), and bringing the nonlinear component ( ) to the right-hand
side of the equation, a pseudo linear system described by Eq. (4.4) is obtained. The left-hand side of
Eq. (4.4) is completely defined by linear terms, whereas the nonlinearities are concentrated in the
pseudo force vector ( ) located in the right-hand side.

( )

( )

( )

(4.4)

Based on Darbre (1996), Fenves and Chavez (1990), Mansur et al. (2000) and Aprile et al. (1994).
28

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

The most consistent way of taking into account the frequency-dependent properties of the reservoirs
compressible fluid, defined by the hydrodynamic force vector (
), is by means of solving the
pseudo linear system described by Eq. (4.4) in the frequency domain. By doing this, in the more
general case, frequency-dependent damping can also be included in the system as shown by Eqs.
(4.5) and (4.6).
and
indicate the constant stiffness and damping matrices, respectively,
whereas
( ) is the frequency dependent damping matrix, ( ) is a general frequency dependent
matrix, and

is the hysteretic damping ratio.

( )

( )

(4.5)
(4.6)

( )

In order to solve Eq. (4.4) in the frequency domain, the excitation force given by the ground
acceleration vector is expressed in the harmonic form shown in Eq. (4.7). Consequently, the harmonic
form of the relative displacement, the hydrodynamic pressure and the pseudo force vectors are
expressed in Eqs. (4.8) to (4.10), respectively.

( )

( )

(4.7)
(4.8)

( )

(4.9)

( )
( )

(4.10)

( )

After replacing Eqs. (4.7) to (4.10) into Eq. (4.4), the exponential common terms are eliminated, and
as a result, the equation of motion of the dam-reservoir system in the frequency domain is defined by
Eq. (4.11) in terms of the complex-valued amplitudes of the ground acceleration [ ( )], the relative
dams displacement [( )], the hydrodynamic pressure [ ( )] and the pseudo force vector [( )].

] ( )

( )

( )

( )

( )

(4.11)

Based on Eq. (3.52), the complex-valued frequency dependent expression of the hydrodynamic
pressure amplitude is determined from the fluid equation of motion given by Eq. (4.12).

(4.12)

Eq. (4.12) is transformed into Eq. (4.13) after expressing the dam absolute acceleration vector ( ) in
terms of the ground ( ) and relative ( ) acceleration vectors.

( )

(4.13)

29

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

After replacing Eqs. (4.7) to (4.9) into Eq. (4.13), the complex-valued frequency dependent
expression of the hydrodynamic pressure amplitude is defined by Eqs. (4.14) and (4.15).

( )

( )

( )

( )

(4.14)

( )]

(4.15)

Eq. (4.16) is obtained by inserting the expression of the hydrodynamic pressure amplitude given by
Eq. (4.14) into the equation of motion of the coupled dam-reservoir system defined by Eq. (4.11). The
hydrodynamic mass matrix ( ) that appears in both sides of Eq. (4.16) is a complex-valued
frequency-dependent term defined by Eq. (4.17), which represents the added mass effect in the
dam provided by the reservoir.

] ( )

( )

( )

(4.16)

( )

(4.17)

( )

Another possibility of expressing the effect of the reservoir on the dam is to split
in its real (added
mass) and imaginary (added damping) components, as shown in Eq. (4.18). The expressions for
the added mass ( ) and damping ( ) are given by Eqs. (4.19) and (4.20), respectively.

] ( )

( )

( )

( )
(4.18)

(4.19)

)
(

(4.20)

The mode superposition method can be employed by introducing the variable defined by Eq. (4.21).
Consequently, the pre-multiplication of Eq. (4.18) by the matrix of modal vectors transposed (
)
results in the equation of motion, defined by Eqs. (4.22) to (4.28), in terms of the generalized modal
coordinates vector ( ).

( )
[

(4.21)

( )
( )

] ( )

( )

( )

(4.22)
(4.23)

( )
( )

(4.24)
(4.25)

30

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

( )

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

(4.26)
(4.27)

( )

(4.28)

The equation of motion in the generalized modal coordinates constitutes a condensed system of
equations in which the unknowns are reduced to a number of chosen modes of vibration. In the
general case only the frequency independent matrices
and
are diagonal, which means that the
system is not uncoupled.
One way of uncouple numerically Eq. (4.22) is to express the full matrices
,
and
as the
summation of a diagonal (d) and an off-diagonal (od) terms as shown in Eqs. (4.29) to (4.31).
(4.29)
(4.30)
(4.31)

Replacing Eqs. (4.29) to (4.31) into Eq. (4.22), and bringing the off-diagonal terms to the right-hand
side, results in an uncoupled system of equations defined by Eq. (4.32) which matrices in the lefthand side are diagonal. The uncoupled system can be solved iteratively by treating the off-diagonal
terms as a second set of pseudo forces dependent of the generalized modal coordinates vector and
defined by Eq. (4.33).

( )

( )

( )

( )

( )
(4.32)

( )

(4.33)

Nevertheless, the introduction of additional pseudo-force vectors to impose the uncoupling of the
system of equations could lead to a cumbersome numerical process, especially in large models with
high amount of DOF.

31

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

4.3. Implementation issues3

4.3.1. Solution procedure sequence of the HFTD method


The HFTD method can be summarized in the following general steps, graphically schematized in the
flowchart of Fig. 4.1.
1. Select the number of modes to be included in the response calculation.
2. Do the eigenvalue analysis of the dam with empty reservoir and calculate the generalized
mass (BS), damping (DS) and stiffness (ES) matrices.
3. Determine the range of excitation frequencies to be used in the solution of the linearized
equation [Eq. (4.22)].
4. Determine for each excitation frequency the generalized added mass (BF), added damping
(DF) and load (Fg). If applicable, the frequency dependent part of the generalized
structural damping matrix (DS) should be calculated for each excitation frequency.
5. Initialize the pseudo force vector in time domain [ ( )]. Only for the first iteration, the
pseudo force vector should be set equal to zero.
6. Perform the Fourier transformation of the ground acceleration signal from the time
domain [ ( )] to the frequency domain [ ( )].
7. Set up the current generalized pseudo force vector in the time domain [
( )].
8. Perform the Fourier transform of the generalized pseudo force vector to the frequency
( )].
domain [
9. Solve the linearized Eq. (4.22) for each excitation frequency and determine the
generalized displacement in the frequency domain [ ( )].
10. Perform the inverse Fourier transform of the generalized displacement ( ), velocity ( )
and acceleration ( ) to the time domain.
11. Expand the generalized response in the time domain to the natural system of coordinates
( , and ) using Eq. (4.21).
12. Perform the state determination in the time domain: Calculation of the nonlinear internal
force of the system using the appropriate constitutive law and the response obtained in
step 11.
) based on the response calculated in
13. Determine the linear restoring force (
step 11.
14. Evaluate the new pseudo force vector using Eq. (4.3) and the results obtained in steps 12
and 13.
15. Check the convergence of the displacement and pseudo force. If convergence is not
achieved, go to step 7 and perform a new iteration.

Based on Darbre and Wolf (1988), Fenves and Chavez (1990), Darbre (1990), Chavez and Fenves (1993),
Chavez and Fenves (1994) and Veletsos and Ventura (1985).
32

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

START

Eigen-analysis and calculation of the generalized structural matrices

Calculation of the generalized added mass [ ( )], added damping [


load [ ( )]

)] and

Initialization of pseudo force vector ( )

Fourier transform of the ground acceleration signal ( )

( )

Calculate the current generalized pseudo force vector

Fourier transform of the generalized pseudo force vector

( )

()

( )

Solve the pseudo linear Eq. (4.22) and determine the generalized
response in the frequency domain ( ), ( ), ( )

Inverse Fourier transform of the generalized response


( )
( ), ( )
( ), ( )
( )

Expand the generalized response to the natural system of coordinates


( )
( ), ( )
( ), ( )
( )

Calculate the new pseudo force vector


(
( )

NO

Convergence?
YES
FINISH

Figure 4.1: Flowchart of the HFTD method general solution procedure

33

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

4.3.2. HFTD method time segmentation approach


The HFTD procedure converges in a time progressive manner, which means that convergence at any
time is reached only after the solution has converged at all previous times (Darbre and Wolf, 1988).
If the procedure described above is executed directly along the whole time span of interest,
convergence and stability problems are likely to occur.
For this reason, the HFTD method is performed following a time segmentation approach in which the
displacements and pseudo forces are evaluated for a limited number of consecutive time steps
contained in time segments. Once convergence is achieved in all the time steps within a segment, the
iterative procedure is repeated for the next time segment, until the total time span of interest is
covered.
The minimum number of time segments required depends on stability conditions4. When a system is
highly stable, the number of time segments and iterations required to obtain convergence diminish.
Nevertheless, the minimum number of iterations required to obtain convergence is more sensitive to
the degree of stability than the minimum number of time segments. This means that two different
systems can converge with the same number of time segments, but the system which is less stable
will always require a higher number of iterations.
In conclusion, the time span of interest has to be divided into time segments containing one or
several time steps. It is important to mention that the length of time segments does not affect the
accuracy of the results obtained but leads to convergent (small length) or divergent (very large
length) solution. When stability or convergence problems occur, then the number of time segments
should be increased by reducing their size (less number of time steps contained in one segment). The
extreme case is given by segments consisting in just one time step.
On the other hand, the number of iterations and the computational effort are proportional to the
number of segments used. If many segments consisting of a few time step each are defined for the
iterative procedure, then the computational effort and the total number of iterations required to
achieve convergence is high. However, this proportional dependence is not critic unless a considerably
higher number of segments than the optimal are used.

4.3.3. Fourier transform period


The period of time used in the Discrete Fourier Transform (Section 2.4) indicates the period of the
periodic loading function (ground acceleration or pseudo force) acting on the system. Therefore, if the
loading (earthquake) duration (t0) is used to transform the ground acceleration and pseudo force to
the frequency domain, the results obtained from the HFTD method would be the steady-state
response of the system subject to a periodic loading with a period equal to to, which harmonic
components are given by its Fourier transform.
The purpose of the HFTD method is to determine the actual transient response of the system subject
to an arbitrary ground acceleration loading, and not the steady-state response of the system subject
to a periodic loading. Therefore, to obtain an accurate approximation of the transient response, the
period used in the Fourier transform should be extended sufficiently long that the effects of the
steady-state response and the loading periodicity become negligible (Veletsos and Ventura, 1985).
4

For a detailed discussion of HFTD method stability criteria see Darbre and Wolf (1988).
34

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

The Fourier period (Tp) should be defined as the summation of the ground acceleration duration (t o)
plus an additional extended period called quiet zone (Tq). The quiet zone is a band of zeros added at
the end of the ground acceleration signal, which length depends on the free vibration properties of
the system (fundamental period of vibration and viscous damping ratio of the structure).
Steps 6, 8 and 10 of the HFTD solution procedure outlined in Section 4.3.1 indicate that the Fourier
transform is applied to the earthquake ground acceleration (direct), the pseudo force (direct) and the
response of the system (inverse), respectively. All these transformations should always be performed
over the same Fourier period (Tp), which is calculated using Eqs. (4.36) and (4.37), in terms of the
fundamental vibration period ( ) and the viscous damping ratio of the system ( ) (Chavez and
Fenves, 1984).
(4.36)
(4.37)

Eq. (4.37) shows that for long oscillation periods (low frequency systems) and low damping ratio, the
quiet zone (Tq) is longer. In the extreme case of a damping ratio equal to zero Tq tends to infinity,
which means that the HFTD method is not applicable for undamped systems (Veletsos and Ventura,
1985).
As mentioned before, the quiet zone is necessary to satisfy the initial conditions and consequently
obtain a better approximation of the transient response instead of the steady state response. On the
contrary, the use of Fourier transformation period which is too short provides inaccurate results, no
matter that they be stable and apparently reasonable (Darbre and Wolf, 1988). In conclusion, the
insertion of a quiet zone should be understood as a numerical representation of what theoretically
should be an infinity period in the analytic Fourier integral (See Section 2.4).
Fig. 4.2 shows a ground acceleration signal which Fourier Transform is calculated twice, one using a
Fourier period equal to the ground acceleration duration (
), and the other using an
extended Fourier period that includes the quiet zone (

).

Figure 4.2: El Centro N-S earthquake ground acceleration signal.

35

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

The spectrums of the ground acceleration amplitudes obtained in both cases are shown in Figs. 4.3
and 4.4, respectively. It can be noticed that both spectrums have similar shapes, with the difference
that the one obtained with the extended Fourier period (Fig. 4.4) is smoother due the higher number
of points used in its definition.
The extended period is ten times the loading duration, and therefore, the number of points used in
both cases holds the same relation. This is the reason why the magnitudes of the ground acceleration
amplitudes of Fig. 4.4 are around ten times the amplitudes of Fig. 4.3.
In order to obtain the transient response of the system, the HFTD method uses the ground
acceleration spectrum shown in Fig. 4.4.

Figure 4.3: Spectrum of the ground acceleration amplitude for Fourier period equal to the
loading duration (
)

Figure 4.4: Spectrum of the ground acceleration amplitude for extended Fourier period
including quiet zone (
)

36

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.3.4. Details of the HFTD


segmentation approach

method

4.

solution

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

procedure

using

the

time

Using a time segmentation approach implies the repetition of the HFTD solution outlined in Section
4.3.1 (from step 6 to step 15) for all the segments defined in the time span of interest. It is important
to realize that the time span of interest (T r) is not the same as the Fourier period (Tp), but the time
for which the calculation of the system response is desired.
Logically, the time span of interest (Tr) is shorter than the period used for the Fourier transform (T p).
As shown in Eq. (4.38), the minimum value of Tr includes the ground acceleration duration (t o) plus
and additional free vibration time equal to one half of the fundamental period of vibration, during
which a maximum peak response could still be attained. The time segmentation should be performed
only over Tr, and the HFTD solution procedure is concluded when the last segment of T r has
converged.
(4.38)

Fig. 4.5 shows the kth time segment and the time steps contained inside. The initial points of time
segments k (Tk) and k+1 (Tk+1), the last converged time step (tc) of the previous iteration, the
time step size (t), and the time segment size (Ts) can also be identified from Fig. 4.5. The initial
time step of segment k is calculated with Eq. (4.39).

(4.39)

Figure 4.5: Graphic representation of the kth time segment.

The HFTD procedure outlined in Section 4.3.1 indicates that, for the first iteration of the first time
segment only, the pseudo force vector is set equal to zero (step 5) and transformed to the frequency
domain (step 8) using the Fourier transformation period (Tp). Once the pseudo linear system has been
solved in the frequency domain (step 9), the generalized response is transformed back to the time
domain (step 10).

37

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

For the subsequent iterations, the segmentation approach requires that the pseudo forces and
displacements in the time domain are calculated and updated (step 11 to step 14) only for the not
converged time steps belonging to the current time segment (tc < t Tk+1 - t). The iterative process
(step 7 to step 14) is repeated until the convergence criteria (step 15) are fulfilled for all the time
steps inside the current segment of analysis (tc = Tk+1 - t).
When all the time steps inside the current segment have converged (tc = Tk+1 - t), the next segment
is investigated. However, in the first iteration of the new segment (step 6) the values of the pseudo
forces belonging to the previous converged time segments are maintained (the pseudo force values of
the converged previous time steps are not modified nor updated anymore), whereas for the current
and subsequent segments, the values of the pseudo forces are set to zero (the definition of a
decaying appended function is also possible as explained in the next section).
In this way, the iterative procedure is continued until the last segment of the time span of interest is
covered, doing always the Fourier transformation through all the period T p, but updating progressively
the displacement and pseudo force vectors only for the non-converged time steps inside the current
segment of analysis. The modified version of the HFTD method flowchart, including time
segmentation approach in the solution procedure, is shown in Fig. 4.6
As a final point, special attention should be given to the fact that the equation of motion of the
pseudo linear system (not the real non-linear system) is the one solved in the frequency domain.
Therefore, the eigenvalues, eigenvectors and critical damping ratios, required to perform the mode
superposition method and to determine the length of the Fourier transformation period (T p) [Eqs.
(4.36) and (4.37)], should be based on the properties of the pseudo linear system defined in Eq.
(4.22).

38

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

START

Eigen-analysis and calculation of the generalized structural matrices

Calculation of the generalized added mass [ ( )], added damping [


load [ ( )]

)] and

Initialization of pseudo force vector ( )

Fourier transform of the ground acceleration signal ( )

( )

Calculate the current generalized pseudo force vector

Fourier transform of the generalized pseudo force vector

( )

()

( )

Solve the pseudo linear Eq. (4.22) and determine the generalized
response in the frequency domain ( ), ( ), ( )

Inverse Fourier transform of the generalized response


( )
( ), ( )
( ), ( )
( )

Expand the generalized response to the natural system of coordinates


( )
( ), ( )
( ), ( )
( )

Calculate the new pseudo force vector


(
( )

NO

Convergence?
YES

YES

Next time segment?


NO
FINISH

Figure 4.6: Flowchart of the modified HFTD method solution procedure including the time
segmentation approach
39

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

4.3.5. Definition of the appended decaying functions for the ground acceleration
loading and the pseudo force
Due to the segmentation approach for solving the linearized equation in the frequency domain, both
load vectors in the right-hand side of Eq. (4.22), ground acceleration [ ( )] and pseudo force
[ ( )], are filled with zeros for all the time steps belonging to the time segments coming after the
current. Therefore, the expressions defining ( ) and ( ), for the time segment k and iteration
j, are given by Eqs. (4.40) and (4.41), respectively, in terms of the Heaviside function (H).

( )

( )[ ( )

( )

(4.40)

)]

( )[ ( )

)]

(4.41)

This means that both time load vectors die out suddenly between the time steps T k+1 t and
Tk+1. In order to avoid this sudden unloading and consequent inaccurate results, a decaying function
may be appended to both load components. The decaying function must start in the last point of the
current time segment (k+1 t) and its extension (Td) is recommended not to be longer than one
segment length (Ts), due to implementation more than performance or accuracy reasons (Darbre,
1990).
One type of decaying function that can be used for this purpose is the sinusoidal function proposed by
Darbre (1990), which is defined by Eqs. (4.42) and (4.43). This decaying function matches the value,
slope and curvature of the loading signal in the time step k+1 t.

( )

[ ( )

( )

( )] [

(4.42)
(4.43)

As expressed in Eq. (4.40), the ground acceleration signal is taken from 0 to Tk+1 t, neglecting
the subsequent part. Instead, the decaying function [Eq. (4.42)] is appended to the signal in the time
point Tk+1 t, followed by a band of zeros until the Fourier period (Tp) is completed.
Therefore, the modified ground acceleration signal for the time segment k is expressed in Eq.
(4.44), where the definition of
is given by Eq. (4.45). Fig. 4.7 is a graphical representation of Eq.
(4.44) applied to the ground acceleration loading shown in Fig. 4.2.

( )

( )[ ( )

)]

( )

(4.44)
(4.45)

40

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

Figure 4.7: Modified excitation load with appended decaying function for the HFTD
analysis of the kth time segment.

On the other hand, for the first iteration the pseudo force vector is set equal to zero for all the time
steps between 0 and Tp. For the next iterations, the values of the pseudo force vector corresponding
to the converged time steps (from 0 to tc) are kept fixed, whereas for the other not converged time
steps inside the current segment (from tc to Tk+1 t), the pseudo force values corresponding to the
last iteration are assigned.
Similarly as in the case of the earthquake ground acceleration, the pseudo force values of the
following time segments correspond to the appended decaying function, continuing with a band of
zeros until completing the Fourier period (Tp). The modified pseudo force vector for the time segment
k and iteration j is expressed in Eq. (4.46).

( )

( )[ ( )

)]

( )

(4.46)

4.3.6. Convergence criteria


The convergence criteria check is performed for both, the pseudo force vector and the displacement
vector in natural or generalized coordinates. For each time step inside the current segment, the
convergence criteria verification after iteration j is given by Eqs. (4.47) and (4.48).
|

( )
|

( )|
( )|

( )
|

( )|
( )|

(4.47)
(4.48)

41

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

4.4. HFTD method input parameters5

4.4.1. Number of modes included in the mode superposition method


The number of modes to be included in the analysis can be determined from the percentage of
effective mass in the direction of analysis, which is obtained from the eigen-analysis of the system
(Section 3.1.9). The general criterion is to include the minimum number of modes which effective
masses in the direction of analysis provide 90% of the total mass of the structure 6.
The effective masses in the three translational directions of analysis corresponding to each mode of
vibration are part of the output results of the eigen-analysis. For example, the effective mass in the Xdirection for the mode of vibration i is defined by Eqs. (4.49), (4.50) and (4.51).
(4.49)
( )

(4.50)
(4.51)

4.4.2. Range of excitation frequencies for the frequency domain solution


The range of frequency samples used in the Fourier analysis goes from 0 to F, which is the maximum
sampling frequency. Taking into account that the Discrete Fourier Transformation (DFT) considers
only positive frequency samples (one-sided expansion), the maximum signal frequency effectively
processed is one half of the maximum sampling frequency (Section 2.4). This relation between the
highest harmonic component of the signal included in the analysis (Nyquist frequency) and the
maximum sampling frequency (F), expressed in Eq. (4.52), is called the Nyquist sampling theorem
(Craig and Kurdila, 2006).
(4.52)

The value of the maximum sampling frequency (F) is inversely proportional to the time step size (t).
This means that once t is defined, F is determined using Eq. (4.53).

(4.53)

Based on Chavez and Fenves (1994), Veletsos and Ventura (1985), Craig and Kurdila (2006) and TNO DIANA
BV (2011).
6
In case of analysis including dam-foundation interaction, the mass of the soil should not be taken into account
in the calculation of the effective mass.
42

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

4.4.3. Time step size


The time step size should be small enough when compared to the more significant oscillation periods
of the loading harmonic components and to the fundamental free vibration period of the structural
system. The first condition ensures the correct representation of the forced or steady-state component
of the systems total response, whereas the second condition allows obtaining a correct representation
of the transient free vibration component (Veletsos and Ventura, 1985).
If high frequency systems subject to long duration loading are analyzed, then the first condition based
on the loading period of oscillation rules the size of the time step. On the other hand, in the case of
low frequency systems subject to short duration loading (f.e. earthquake analysis of high dams), the
second condition based on the fundamental free vibration period is the most relevant criterion.
The time step size which takes into account both criteria fulfills the conditions stated in Eq. (4.54), in
terms of the fundamental vibration frequency (F1), the vibration frequency of the highest mode
included in the analysis (Fq), and the greater frequency usually contained in earthquake ground
acceleration records (25 Hz) (Chavez and Fenves, 1994).

(4.54)

When the problem to be solved with the HFTD method is highly nonlinear (f.e. problems including
contact or base sliding nonlinearities), it is recommended that in addition the time step size fulfills Eq.
(4.55) (Chavez and Fenves, 1994).
(4.55)

4.4.4. Parameters for the Discrete Fourier Transform (DFT)


As explained before, the total period for the Fourier Transform (T p) is calculated using Eqs. (4.36) and
(4.37). Additionally, it is recommended that Tp fulfills the condition stated by Eq. (4.56) (Chavez and
Fenves, 1994). However the very long value of Tp obtained with Eq. (4.56) may not be manageable,
in terms of computational time and storage memory capacity, for large multi degree of freedom
(MDOF) systems. Then in those cases the use of a reduced Fourier period should be evaluated
without prejudice of the accuracy of the results obtained.
(4.56)

The number of time point samples (N) given by Eq. (4.57) is obtained from the Fourier period (Tp)
and the time step size (t). The frequency increment ( ) is calculated using Eq. (4.58), whereas the
maximum sampling frequency (F) is calculated with Eq. (4.59), which is equivalent to Eq. (4.53).
(4.57)

43

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

4.

Hybrid Frequency-Time Domain (HFTD)


Method Solution Procedure

(4.58)
(4.59)

44

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

5. SINGLE DEGREE OF FREEDOM (SDOF) DYNAMIC SYSTEMS

5.1 Solution procedure of the HFTD method for SDOF non-linear systems
The HFTD method, which was previously explained in detail for seismic analysis of dam-reservoir
dynamic interaction multi degree of freedom (MDOF) systems (Chapter 4), is developed here for the
particular case of SDOF dynamic linear and nonlinear systems.
It was previously stated that the HFTD method was developed to analyze systems with frequency
dependent properties and nonlinear behavior. Therefore, the formulation of the HFTD method for
SDOF systems also considers the inclusion of frequency dependent terms, no matter that in most of
the SDOF numerical examples developed in Section 5.3, these frequency dependent systems are not
taken into account. This is because to evaluate the soundness, precision and stability of the HFTD
method, the results obtained for the SDOF systems have to be compared with other analytical
solutions or established numerical methods (f.e.: Newmark method) in the time domain, which are
naturally capable of dealing with nonlinearities, but not with frequency dependence.
The SDOF system considered is expressed by Eq. (5.1). As it can be noticed, this equation of motion
constitutes a general case with nonlinear behavior [
] and frequency dependent terms
[

]. Therefore, it is not possible to directly solve Eq. (5.1) in the time domain.

(5.1)

Assuming that additive decomposition between linear and nonlinear terms is possible, the internal
nonlinear restoring force is expressed by Eq. (5.2) in terms of the pseudo force [
], the constant
stiffness ( ) and the constant damping ( ) (Aprile et al., 1994).

(5.2)

After replacing Eq. (5.2) into Eq. (5.1) and passing the pseudo force to the right hand side, Eq. (5.3)
is obtained. This equation of motion contains only linear terms in the left hand side, whereas all
sources of nonlinearities are concentrated in the pseudo force term located in the right hand side of
Eq. (5.3).

(5.3)

Obviously, the value of the pseudo force is not known a priori, and for that reason Eq. (5.3) (known
as pseudo linear or linearized equation) has to be solved iteratively until convergence is achieved for
both, the displacement solution ( ) and the pseudo force ( ). Based on Eq. (5.2), Eq. (5.4) shows the
expression to calculate the value of the pseudo force in iteration j.

(5.4)
45

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

If there were no frequency dependent terms, the pseudo linear equation could be solved in both,
frequency and time domains. However, frequency dependent terms are considered in Eq. (5.3) and
therefore, the HFTD method solves the pseudo linear system in the frequency domain.
Accordingly, the excitation force is expressed in terms of its harmonic components defined in Eq.
(5.5). As a consequence, the displacement and the pseudo force are defined in the same terms by
Eqs. (5.6) and (5.7), respectively. N is the number of excitation frequencies ( ) contained in the
force function (or signal)
.

(5.5)

(5.6)

(5.7)

Eq. (5.8) is the pseudo linear equation of motion in the frequency domain, which is solved for each
excitation frequency ( ). Eq. (5.8) is obtained in terms of the complex valued amplitudes of the
excitation force [
], displacement [
] and pseudo force [
], after replacing Eqs. (5.5)
to (5.7) into Eq. (5.3).

(5.8)

Prior to solve Eq. (5.8), the complex right hand side amplitudes [

] have to be
determined. Then, the equation is solved and the complex amplitude of the displacement [
] is
obtained. The HFTD method uses the Discrete Fourier Transform (DFT) defined in Eqs. (5.9) and
(5.10) to determine
and
, respectively.

(5.9)
(5.10)

The number of excitation frequencies (N) considered in the analysis depends on the period used for
the Fourier transform (Tp) and the time step size (t) that determines the separation of the discrete
time points in which the response of the system is computed. Eq. (5.11) presents the expression to
determine N, whereas Eq. (5.12) the expression to calculate
.

(5.11)

(5.12)

46

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

In Chapter 4 it was explained that in the HFTD method the Fourier period (T p) is not equal to but
longer than the duration of the excitation force (t0). The reason for this is to obtain the transient
response of the system instead of the steady state response (Section 4.3.3). The expression to
calculate the ideal value of the Fourier period is given by Eq. (5.13), where the natural period of the
undamped system (T1) and the viscous damping ratio ( ) are defined by Eqs. (5.14) and (5.15),
respectively. For the SDOF systems treated in this chapter, the very long value of Tp obtained with Eq.
(5.13) is totally manageable in terms of computational time and memory capacity.

(5.14)

(5.13)

(5.15)

The fundaments and procedure of the time segmentation approach followed by the HFTD method to
achieve convergence when solving nonlinear systems have already been explained in Section 4.3. This
procedure consists in dividing the time span of interest (not the total Fourier period) in time segments
over which the iterative solution is found sequentially for each of these segments.
The time span of interest (Tr) is the time for which the user desires to obtain the response of the
system. Usually, Tr is equal or longer than the load duration but much smaller than the Fourier period
(t0 Tr << Tp).
On the other hand, all the time segments are generally composed of a constant number of time steps.
Therefore, the size of each time segment (TS) is calculated with Eq. (5.16), in terms of the number
of samples (or time points) per segment (N S) and the time step size (t), whereas the starting time
point of segment k (Tk) is calculated using Eq. (5.17).
(5.16)
(5.17)

Fig. 5.1 shows an example of the excitation load


with duration equal to t0 and Fourier period
equal to Tp. In this particular case the time span of interest (Tr) is divided in six time segments. For
each time segment,
and
should be calculated by means of the DFT expressed in Eqs.
(5.9) and (5.10), respectively.
If the kth time segment of a system subjected to the load shown in Fig. 5.1 is being analyzed, the
segmentation approach of the HFTD method requires modifying the excitation load
before doing
the DFT. The required modifications of load
are shown in Fig. 5.2. The original force signal is
replicated until the kth segment currently analyzed. Subsequently, an appended decaying function that
develops a smooth transition to load values equal to zero is added. Finally, the zero load values are
extended until covering all the time points contained by the Fourier period (Tp).
47

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.1: Excitation load, segmentation approach, time span of interest and Fourier
period.

Figure 5.2: Modified excitation load with appended decaying function for the HFTD
analysis of the kth time segment.

The use of a decaying function avoids a sudden unloading of the system, and consequent inaccuracy
and artificial oscillations in the obtained response. Taking into account that in the pseudo linear
equation of motion solved by the HFTD method, the frequency dependent complex amplitude terms
located in the right hand side of Eq. (5.8) do not only consist in the Fourier transform of the load
48

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

[
], but of the pseudo force [
] as well, a similar modification shown in Fig. 5.2 for
be accomplished for
before the DFT defined in Eq. (5.10) is applied.

has to

The modifications of the excitation load and the pseudo force for the analysis of the kth time segment
are analytically expressed in Eqs. (5.18) and (5.19), respectively, by means of the Heaviside function
[
].
(5.18)
(5.19)

Several types of decaying functions can be used for the load [

] and pseudo force [

].

Nevertheless, the simplest decaying function defined by Eqs. (5.20) and (5.21) is selected for both of
them. This is a linear decaying function defined in terms of its time variable ( ) and its duration
(
).
starts in the last point of the time segment k, whereas by definition
should be smaller
than the time segment length (
).

(5.20)

(5.21)

Going back to the solution of Eq. (5.8), the complex amplitude of the displacement ( ) is determined
in the frequency domain for each excitation frequency ( ). The complex amplitudes of the systems
velocity and acceleration in the frequency domain are determined for each excitation frequency using
Eqs. (5.22) and (5.23).

(5.22)

(5.23)

From the amplitudes of the system response in the frequency domain, the response in the time
domain is obtained by means of the inverse DFT, expressed in Eqs. (5.24) to (5.26) for the
displacement, velocity and acceleration, respectively.

(5.24)

(5.25)

(5.26)

49

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Due to the segmentation approach followed in the HFTD method, the response in the time domain is
only required to be calculated for those time points located inside the current segment being
analyzed, for which convergence has not been achieved yet. This means that if the last time point in
the segment k for which convergence has been achieved in the previous iteration is denominated ,
then the time points
, for which Eqs. (5.24) to (5.26) have to be evaluated, fulfill the condition
expressed in Eq. (5.27).
(5.27)

Finally, with the response obtained for the time points

, the pseudo force for iteration j is obtained

after replacing
and
in Eq. (5.4). Then, the convergence criteria check is performed for
both, the pseudo force and the displacement. The convergence criteria verification for iteration j is
presented in Eqs. (5.28) and (5.29).
|
|

(5.28)

(5.29)

If one of the time points


does not fulfill the convergence criteria defined by Eqs. (5.28) and (5.29),
then a new iteration is initiated with the DFT of the pseudo force obtained in the previous iteration.
The last time point for which convergence has been achieved ( ) should be updated, being the
process repeated until
coincides with the time span of interest (
). Tables 5.1 and 5.2 show a
schematic summary of the HFTD method for SDOF non-linear systems.

50

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Table 5.1: Summary of the HFTD method for SDOF non-linear systems
1.

Initial calculations
1.1

Select the time step size (t)

1.2

Determine the duration of the excitation load ( )

1.3

Calculate the fundamental period of the system ( )


1.4

Calculate the damping ratio ( )

2.

Fourier period
2.1

Calculate the Fourier period ( )


{

2.2

Calculate the number of time points ( ) for the Fourier transform

3.

4.

Frequency dependent properties


3.1

Calculate the excitation frequencies (

3.2

Evaluate the frequency dependent damping [

] for each excitation frequency

Number of time segments


4.1

Select the time span of interest ( )

4.2

Select the number of samples (time points) per segment (

4.3

Calculate the size of the time segment (

4.4

Calculate the number of time segments (

5.

Set up the initial the value of the pseudo force equal to zero

6.

Calculations for each time segment k


6.1

Determine the modified excitation load for time segment k (

6.2

Calculate the DFT of the excitation load [

6.3

]
[

Follow the HFTD iterative procedure to determine the system response and pseudo forces of the
time points of segment k (See Table 5.2)

51

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Table 5.2: Iterative procedure of the HFTD method for SDOF nonlinear systems
1.

Calculations for each iteration j


1.1

Determine the modified pseudo force for time segment k and iteration j-1 [

1.2

Calculate the DFT of the pseudo force [

1.3

Determine the modified pseudo force for time segment k and iteration j-1 [

1.4

Calculate the DFT of the pseudo force [

1.5

Calculation for each excitation frequency


1.5.1

Solve the pseudo linear equation to determine displacement amplitude in the frequency
domain
[

1.5.2

Calculate the velocity [

] and acceleration [

] amplitudes in the frequency

domain

1.6

Calculate the inverse DFT of the displacement, velocity and acceleration


[

1.7

Calculation of the internal force [

1.8

Calculation of the pseudo force for the time points

)] for the time points

between

1.9

Convergence criteria verification for the time points


|

|
|

1.10

Update

1.11

Repeat the iterative process until

between

and

and

)
between

and

|
|

5.2 Time domain analysis of SDOF systems


The accuracy of the results obtained with the HFTD method previously described for SDOF systems
has to be compared with analytical solutions or other well established methods of analysis in the time
52

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

domain. This permits to evaluate the HFTD method performance and generate confidence about the
results obtained when frequency dependent properties are actually required to be included.
Nevertheless, it is obvious that the numerical examples tested in the section 5.3 cannot have
frequency dependent properties, due to fact that the benchmark solutions are obtained from time
domain analysis methods.
This section is divided in two parts. The first part covers the analytic solution in the time domain of
SDOF linear systems. Two particular types of loading are considered: harmonic and general. The
second part develops the well-established Newmark method in the time domain for SDOF non-linear
systems.

5.2.1 Analytic solution in the time domain for SDOF linear systems1

Figure 5.3: Single degree of freedom damped system

The equation of motion of the damped SDOF linear system shown in Fig. 5.3 is expressed by Eq.
(5.30) in terms of the relative displacement ( ). The initial conditions expressed in Eqs. (5.31) and
(5.32) complete the definition of the governing equations.

(5.30)
(5.31)

Based on Craig and Kurdila (2006) and Paz and Leigh (2004).

(5.32)

53

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Eq. (5.30) can be expressed as shown in Eq. (5.33). The definition of the natural frequency of
vibration ( ) and the viscous damping ratio ( ) is presented in Eqs. (5.34) and (5.15), respectively.

(5.33)

(5.34)

If the right hand side of Eq. (5.33) is denominated

and the subscript r of relative movement is

omitted, a general equation valid for relative and absolute coordinates is expressed in Eq. (5.35).
(5.35)

5.2.1.1

Response of SDOF systems subjected to harmonic loading

When the excitation force acting in the right-hand side of Eq. (5.35) is harmonic in the form
, then Eq. (5.35) can be expressed as shown in Eq. (5.36).

(5.36)

The solution of Eq. (5.7) is given by the summation of the general homogeneous solution and the
particular solution. For under-damped systems (
), the general solutions is defined by Eqs. (5.37)
and (5.38).
(5.37)
(5.38)

On the other hand, the particular solution or steady state response is given by Eq. (5.39) in terms of
the displacement amplitude ( ) and the phase angle ( ), defined by Eqs. (5.40) and (5.41),
respectively.
(5.39)
(5.40)

) ]

)]

(5.41)

54

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Therefore, the total solution of the SDOF under-damped linear system subjected to the harmonic load
is given by the sum of Eqs. (5.37) and (5.39), as expressed in Eq. (5.42). The derivative
of Eq. (5.42) provides the expression of the velocity of the system, shown in Eq. (5.43). The
acceleration can be obtained from the equation of motion, after the substitution of Eqs. (5.42) and
(5.43) into Eq. (5.36). The values of
and
are obtained from the initial conditions defined in Eqs.
(5.31) and (5.32).
(5.42)

(5.43)

5.2.1.2
Response of SDOF systems subjected to general loading by the
Convolution Integral method
If the loading is not harmonic but a general dynamic excitation
, the particular solution is
obtained by the Convolution (or Duhamel) integral method, which is expressed in Eq. (5.44). The total
solution is given therefore by Eq. (5.45).

(5.44)

(5.45)

5.2.1.3
Response of SDOF systems subjected to general loading by the
Direct Integration method
In many applications, the excitation force
cannot be expressed by a mathematical function (f.e:
ground acceleration signal records) and therefore, the convolution integral of Eq. (5.45) cannot be
obtained in a continuous closed form.
One way to deal with these kind of problems without losing accuracy and theoretical consistency is to
evaluate the convolution integral in a discrete form by dividing the loading in small linear segments, as
shown in Fig. 5.4.
In this way, if the time step size (t) that determine the position of the discrete points in the force
function is small enough, the solution of the system can be calculated for the time step i, using as
initial conditions the solution obtained for the time step i-1. This approach is known as the Direct
Integration Method and provides the exact analytical solution of the system, provided that the time
step size is small enough to describe accurately the force function.

55

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.4: Discretization of the force the function

The force for the time interval between ti and ti+1 is approximated to a linear function by Eqs. (5.46)
and (5.47). Therefore, the equation of motion (only valid for the time interval between t i and ti+1) is
expressed in Eq. (5.48).

(5.46)

( )

(5.47)

[(

( )

(5.48)

The general and particular solutions of the system for the interval i are given by Eqs. (5.49) and
(5.50), respectively. The constants Ci and Di of the general solution are obtained based on the initial
conditions of the interval (i.e.
and ). On the other hand, the constants Ai and Bi are obtained
after replacing the particular solution in the equation of motion defined in Eq. (5.48).
(5.49)
(5.50)

The total solution for displacement and velocity is given respectively by Eqs. (5.51) and (5.52),
respectively.

(5.51)
(5.52)
56

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

After replacing Eq. (5.50) into Eq. (5.48), the expressions defined in Eqs. (5.53) and (5.54) are the
solution for Ai and Bi, respectively.
(5.53)
(5.54)

At the beginning of interval i, the initial conditions are


and . After replacing both conditions in
Eqs. (5.51) and (5.52), respectively, the constants C i and Di are obtained with the expressions given
by Eqs. (5.55) and (5.56).
(5.55)

(5.56)

Finally, the solution of interval i is completely defined and therefore it is possible to determine the
response of the next time step i+1, by replacing the time
in Eqs. (5.51) and (5.52), to obtain
and
respectively. The acceleration (
) is obtained from Eq. (5.48).
After substitution and simplification, the following final recurrence formulas for displacement, velocity
and acceleration of the time step i+1 are presented in Eqs. (5.57) to (5.59). The expressions to
calculate the coefficients of the recurrence formulas are summarized in Table 5.3.
(5.57)

(5.58)
(5.59)

57

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Table 5.3: Coefficients for the recurrence formulas of the Direct Integration Method

[(

5.2.2 Solution procedure of the Newmark method for SDOF non-linear systems2
The equation of motion for nonlinear systems solved by the HFTD method [Eq. (5.1)] is modified in
such a way that no frequency dependent terms are taken into account, as it is expressed in Eq.
(5.60). This general equation for nonlinear SDOF systems is valid for absolute and relative
coordinates, and therefore, the force vector
can include a ground acceleration excitation.

(5.60)

One of the most stable and accurate time domain methods for nonlinear analysis of structural
dynamics systems is the Newmark method. Due to implementation purposes of Newmark method, it is
desirable to split the nonlinear internal force [
] in stiffness [
] and damping
[
] nonlinear forces. This is expressed in Eq. (5.61).

(5.61)

Eq. (5.61) can be discretized so it is valid at discrete time instants t i. Therefore, the nonlinear equation
of motion is satisfied in the time instants ti and ti+1, as it is expressed in Eqs. (5.62) and (5.63).

Based on Chopra (2007).

(5.62)

(5.63)

58

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

The difference between Eqs. (5.63) and (5.62) gives the incremental equilibrium equation expressed
in Eq. (5.64), for which the time step is defined by Eq. (5.65)
(5.64)

(5.65)

The incremental stiffness and damping internal restoring forces can be expressed as shown in Eqs.
(5.66) and (5.67), respectively. This approximation is valid for small step sizes (t) and for tangent
stiffness and damping coefficients defined by Eqs. (5.68) and (5.69), respectively.
(5.66)

(5.67)
(5.68)

(5.69)

After replacing Eqs. (5.66) and (5.67) into Eq. (5.64), the incremental equilibrium equation is
reformulated and presented in Eq. (5.70). This equation has the same form of the equation for linear
systems, and this is the reason for splitting the restoring force in Eq. (5.61).

(5.70)

Now, the Newmark method can be applied in the same fashion as it is done for linear systems. The
only differences are that the tangent stiffness and damping have to be updated for each time step,
and that an iterative procedure has to be included to avoid the error accumulation of pure incremental
procedures (modified Newton-Raphson iterative method)3.
The Newmark method approximates the displacement and velocity of the system in the time instant
ti+1 via the recurrence algorithms presented in Eqs. (5.71) and (5.72), respectively. The values of the
parameters and determine the stability and accuracy of the method. The two most used cases due
to their high accuracy are the linear (=1/2, =1/6) and average (=1/2, =1/4) acceleration
methods.

(5.71)
(5.72)

See section 5.7 of Chopra (2007), for a detailed explanation of the modified Newton-Raphson iterative
procedure applied to nonlinear dynamic analysis of SDOF systems.

59

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

After rearranging Eqs. (5.71) and (5.72), the incremental velocity and acceleration are expressed in
Eqs. (5.73) and (5.74), respectively, in terms of the incremental displacement and the response of the
time instant ti.

(5.73)

(5.74)

Eqs. (5.73) and (5.74) are substituted into Eq. (5.70) to obtain the incremental equilibrium equation,
defined by Eqs. (5.75) to (5.77). The only unknown of Eq. (5.75) is the incremental displacement
(
), which is calculated based on the solution of the previous time instant t i, and following the
modified Newton-Raphson iterative procedure in order to obtain accurate results.

(5.75)

(5.76)

(5.77)

The incremental displacement found after solving Eq. (5.75) can be replaced into Eq. (5.73) to find
. Then, the displacement, velocity and acceleration of the time instant t i+1 are calculated using
Eqs. (5.78) to (5.80), respectively.
(5.78)

(5.79)

(5.80)

The internal restoring forces in the time instant t i+1 are calculated incrementally, as it is expressed in
Eqs. (5.81) and (5.82). The definition of the restoring forces increment is given by Eqs. (5.66) and
(5.67).

(5.81)
(5.82)

Table 5.4 shows a schematic summary of the Newmark method for SDOF nonlinear systems, whereas
Table 5.5 shows the procedure of the Newton-Raphson iterative procedure.

60

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Table 5.4: Summary of the Newmark method for SDOF nonlinear systems4
Special cases:
(1) Average acceleration method (

(2) Linear acceleration method (


1.

Initial calculations
1.1

Select the time step size (t)

1.2

Determine the initial acceleration ( )

2.

Calculations for each time step i = 1,2,3,


2.1

Determine the load increment (

2.2

Determine the tangent stiffness (ki) and damping (ci)

2.3

Determine the effective stiffness coefficient ( )

2.4

Determine the effective load increment (

2.5

Determine the displacement increment (

) using modified Newton-Raphson iterative

procedure to solve the following equation (Table 5.5)

2.6

Determine the velocity increment (

2.7

Determine the displacement (

) and velocity (
;

2.8

) of time step i+1

Determine the internal restoring forces (

) of time step i+1

;
2.9

Determine the acceleration of time step i+1 (

Based on Table 5.7.2 of Chopra (2007)

61

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Table 5.5: Summary of the Modified Newton-Raphson iterative procedure


1.

Initial calculations
1.1

Initial estimation of the displacement for the time step i+1 (

1.2

Initial estimation of the nonlinear stiffness (

1.3

Initial residual force for iteration 1 (

) and damping (

) force for time step i+1

2.

Calculations for each iteration j = 1, 2, 3,


2.1

Solve the equation to determine the displacement increment (

) between iteration j and

iteration j-1

2.2

Determine the total displacement of iteration j for time step i+1 (

2.3

Determine the total displacement increment of iteration j between time steps i and i+1 (

2.4

Determine the total velocity increment of iteration j between time steps i and i+1 (

2.5

Determine the total velocity of iteration j for time step i+1 (

2.6

Determine the internal restoring forces of iteration j for time step i+1 (

) based on

the nonlinear relations force, displacement and velocity


2.7

Determine the variation of the external force after iteration j (

{
2.8

Determine the variation of the internal force after iteration j (


(

2.9

Determine the residual force for the next iteration j+1 (

)
)

62

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

5.3 HFTD method numerical test examples for SDOF systems


The HFTD (Tables 5.1 and 5.2), the Direct Integration (Table 5.3) and the Newmark (Tables 5.4 and
5.5) methods presented in Sections 5.1 and 5.2, were programmed in MATLAB 5 (The Mathworks, Inc.,
2012) to be used in the solution of the examples developed in this section.

5.3.1 Test examples for SDOF linear systems


Four examples of SDOF linear systems are solved with the HFTD method and compared with the
analytic solution for harmonic loading or the Direct Integration method for general loading. Each of
these examples represents different types of loading that cover a wide range of possibilities.

5.3.1.1

Harmonic loading6

The SDOF system is subject to the harmonic load


the characteristics of the system is shown in Table 5.6.

. The numerical data defining

Table 5.6: Numerical data of Example 4.2 (Craig and Kurdila, 2006)
k
40

m
0.1

0.2

p0
10

10

t0
2

Based on the information given in Table 5.6, the parameters of the SDOF system shown in Table 5.7
are determined. The analytical solution for harmonic loading developed in section 2.1 is used to solve
this example. Table 5.8 shows the displacement amplitude (U), phase angle ( ) and constants A1 and
A2, which are obtained using Eqs. (5.40) to (5.43), as well as the information shown in Tables 5.6 and
5.7.
Table 5.7: Parameters of the SDOF linear system subject to harmonic loading
Tr
2

T1
0.314

20

19.60

Table 5.8: Analytic solution of SDOF linear system subject to harmonic loading
U
0.322

0.261

-0.311

-0.1059

Finally, Eqs. (5.83) and (5.84) present the analytic displacement and velocity, obtained from Eqs.
(5.42) and (5.43), respectively. The acceleration response is indirectly determined by replacing the
displacement and velocity values into Eq. (5.36).
(5.83)
5
6

The MATLAB routines can be found in Appendix B.


Example 4.2 of Craig and Kurdila (2006).

63

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

(5.84)

On the other hand, Table 5.9 shows the values of the parameters required to perform the HFTD
method schematized in Tables 5.1 and 5.2. It is important to mention that for linear systems the
HFTD method does not require segmentation approach or iterative procedure. For that reason only
one time segment (NT =1) is used in all the examples corresponding to linear systems.
Table 5.9: HFTD parameters for the SDOF linear system subject to harmonic loading
t

Tp

0.02

15.72

786

NT
0.0636

The comparison between the analytic and the HFTD solutions for the displacement, velocity and
acceleration is presented in Figs. 5.5 to 5.7.
For the displacement and velocity responses, the major differences of the HFTD solution with respect
to the analytical solution are found during the first 0.5 s. For the acceleration, this range is extended
until the time 0.75 s. After this initial period, the HFTD response gets stable and matches almost
exactly the analytical solution thenceforth.
The reason for this initial inaccurate range is that the magnitude of the cosine harmonic load does not
start in zero, but in p0 equal to 10. The HFTD is not capable to exactly trace the response generated
by this sudden loading at the initial time t=0. Once the effect of the sudden initial loading is
diminished, the response obtained with the HFTD method gradually matches the analytical solution.

Figure 5.5: Harmonic loading example. Displacement response for Analytic and HFTD
solutions.
64

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.6: Harmonic loading example. Velocity response for Analytic and HFTD solutions.

Figure 5.7: Harmonic loading example. Acceleration response for Analytic and HFTD
solutions.

5.3.1.2

Blast loading7

The SDOF system is subject to the blast load schematized in Fig. 5.8. The numerical data defining the
characteristics of the system is shown in Table 5.10.
7

Example 4.2 of Paz and Leigh (2004).

65

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.8: Blast loading.

Table 5.10: Numerical data of Example 4.2 (Paz and Leigh, 2004)
k
100 000

m
100

0.2

t0
0.06

Based on the information given in Table 5.10, the parameters of the SDOF system shown in Table
5.11 are determined. The recurrence formulas of the Direct Integration method developed in Section
5.2.1.3 are used to calculate the exact analytic response of the SDOF linear system. Table 5.12 shows
the values of the coefficients for the recurrence formulas of the Direct Integration method. They were
calculated based on the information given in Tables 5.10 and 5.11, as well as the expressions of Table
5.3.
Table 5.11: Parameters of the SDOF linear system subject to blast loading
Tr
0.20

T1
0.20

31.62

30.92

Table 5.12: Direct Integration method coefficients for the SDOF linear system subject to
blast loading
t
0.005

6.32

A
8.1x10-8

B
4.1x10-8

C
0.9878

D
0.0048

A
2.4x10-5

B
2.4x10-5

C
-4.825

D
0.9268

On the other hand, Table 5.13 shows the values of the parameters required to perform the HFTD
method (Tables 5.1 and 5.2). Similarly as in the previous example, for linear analysis the
segmentation approach is not required (only one time segment is required to be defined).
Table 5.13: HFTD parameters for the SDOF linear system subject to blast loading
t

Tp

0.005

9.935

1987

NT
0.1007

1
66

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

The comparison between the analytic Direct Integration and the HFTD solutions for the displacement,
velocity and acceleration is presented in Figs. 5.9 to 5.11. The accuracy obtained in the three
responses is remarkable. The variations are so small that are almost no perceptible. This means that
when the loading starts in zero (see Fig. 5.8) the HFTD method shows a great performance.

Figure 5.9: Blast loading example. Displacement response for Direct Integration and HFTD
solutions.

Figure 5.10: Blast loading example. Velocity response for Direct Integration and HFTD
solutions.

67

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.11: Blast loading example. Acceleration response for Direct Integration and
HFTD solutions.

5.3.1.3

Impulsive loading8

The SDOF system is subject to the impulse load schematized in Fig. 5.12. The numerical data defining
the characteristics of the system is shown in Table 5.14.

Figure 5.12: Impulsive loading.

Example 4.7 of Paz and Leigh (2004).

68

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Table 5.14: Numerical data of Example 4.7 (Paz and Leigh, 2004)
k
10 000

m
10

0.1

t0
0.10

Based on the information given in Table 5.14, the parameters of the SDOF system shown in Table
5.15 are determined. Table 5.16 shows the values of the coefficients for the recurrence formulas of
the Direct Integration method [Eqs. (5.57) to (5.59)].
Table 5.15: Parameters of the SDOF linear system subject to impulsive loading
Tr
0.22

T1
0.20

31.62

31.46

Table 5.16: Direct Integration method coefficients for the SDOF linear system subject to
impulsive loading
t
0.01

3.16

A
3.2x10-6

B
1.6x10-6

C
0.9514

D
0.0095

A
4.7x10-4

B
4.9x10-4

C
-9.53

D
0.8912

On the other hand, Table 5.17 shows the values of the parameters required to perform the HFTD
method. Similarly as in the previous example, for linear analysis only one time segment is required.
Table 5.17: HFTD parameters for the SDOF linear system subject to impulsive loading
t

Tp

0.01

9.94

994

NT
0.1006

The comparison between the analytic Direct Integration and the HFTD solutions for the displacement,
velocity and acceleration is presented in Figs. 5.13 to 5.15. The accuracy obtained in the three
responses is as good as in the previous example.

69

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.13: Impulsive loading example. Displacement response for Direct Integration
and HFTD solutions.

Figure 5.14: Impulsive loading example. Velocity response for Direct Integration and
HFTD solutions.

70

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.15: Impulsive loading example. Acceleration response for Direct Integration and
HFTD solutions.

5.3.1.4

Ground acceleration record loading9

The SDOF system is subject to the ground acceleration correspondent to the North-South component
of the El Centro earthquake shown in Fig. 5.16. The numerical data defining the characteristics of the
system is shown in Table 5.18.

Figure 5.16: El Centro N-S earthquake ground acceleration.

Example 4.9 of Paz and Leigh (2004).

71

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Table 5.18: Numerical data of Example 4.9 (Paz and Leigh, 2004)
K

t0

2136

38.86

0.05

4.5

Based on the information given in Table 5.18, the parameters of the SDOF system shown in Table
5.19 are determined. Table 5.20 shows the values of the coefficients for the recurrence formulas of
the Direct Integration method. On the other hand, Table 5.21 shows the values of the parameters
required to perform the HFTD method.
Table 5.19: Parameters of the SDOF linear system subject to earthquake ground
acceleration
Tr
4.38

T1
0.85

7.414

7.405

Table 5.20: Direct Integration method coefficients for the SDOF linear system subject to
earthquake ground acceleration
t
0.02

0.37

A
3.4x10-6

B
1.7x10-6

C
0.9891

D
0.0198

A
2.5x10-4

B
2.5x10-4

C
-1.09

D
0.9744

Table 5.21: HFTD parameters for the SDOF linear system subject to earthquake ground
acceleration
t

Tp

0.02

42.38

2119

NT
0.0236

The comparison between the analytic Direct Integration and the HFTD solutions for the displacement,
velocity and acceleration is presented in Figs. 5.17 to 5.19. The accuracy obtained in the three
responses is as good as in the previous example.

72

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.17: Ground acceleration loading example. Displacement response for Direct
Integration and HFTD solutions.

Figure 5.18: Ground acceleration loading example. Velocity response for Direct
Integration and HFTD solutions.

73

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.19: Ground acceleration loading example. Acceleration response for Direct
Integration and HFTD solutions.

5.3.2 Resonance case study: SDOF linear systems subject to harmonic loading
with exponential amplitude
The SDOF linear system is subject to a harmonic sinusoidal load which amplitude varies exponentially,
as expressed by Eq. (5.85). In order to test the stability of the HFTD method, the influence of load
excitation frequencies close to the resonance condition ( = o) is studied for systems with low
damping ratios ( 0.05)10. In addition, the loading duration (t0) is expressed in Eq. (5.86) as a
number n of load excitation periods (T) or peaks.

(5.85)
(5.86)

The numerical data of the SDOF system to be studied is shown in Table 5.22. As it can be noticed,
three different low damping ratios are taken into account.
Table 5.22: Numerical data of the SDOF system for resonance case studies
k
2136

m
38.86

7.414

(0.01 ; 0.03 ; 0.05)

First, four load excitation frequencies close to the natural vibration frequency of the system are
considered with a damping ratio equal to 0.05. The characteristics of these four case studies are
summarized in Table 5.23.
10

As explained in Section 4.3, the HFTD method is not applicable to undamped systems.

74

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Table 5.23: Numerical data of case study A: Excitation frequencies close to resonance
Case
A-1
A-2
A-3
A-4

0.05
0.05
0.05
0.05

0.80
0.95
1.05
1.20

1.06
0.89
0.81
0.71

2500
2500
2500
2500

25
25
25
25

26.49
22.31
20.19
17.65

The harmonic load


given by Eq. (5.85) is shown in Fig. 5.20 as example for the particular case A4. Fig. 5.20 shows the maximum amplitude , the number of peaks or load periods and the load
duration . Similar graphics can be presented for the other three cases.

Figure 5.20: Harmonic sinusoidal load with exponential amplitude for the case study A-4.

Like in the previous examples, the results obtained with the HFTD method (Tables 5.1 and 5.2) are
compared with the analytical solution provided by the Direct Integration method described in Section
5.2.1.3, and summarized by Eqs. (5.57) to (5.59) and Table 5.3.
The HFTD parameters for each of the four case studies are summarized in Table 5.24. Due to lack of
space, only the results obtained for the displacement response of the SDOF linear system are
presented and compared with the Direct Integration method. The analysis of velocity and acceleration
results revealed as good performance as the displacement results, presented in Figs. 5.21 to 5.24 for
the cases A-1, A-2, A-3 and A-4, respectively.
Table 5.24: HFTD parameters for case study A
Case

Tp

A-1
A-2
A-3
A-4

0.025
0.025
0.025
0.025

53.00
44.68
42.38
42.38

2120
1787
1695
1695

NT
0.0189
0.0224
0.0236
0.0236

1
1
1
1
75

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.21: Resonance case study A-1. Displacement response for Direct Integration and
HFTD solutions.

Figure 5.22: Resonance case study A-2. Displacement response for Direct Integration and
HFTD solutions.

76

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.23: Resonance case study A-3. Displacement response for Direct Integration and
HFTD solutions.

Figure 5.24: Resonance case study A-4. Displacement response for Direct Integration and
HFTD solutions.

From the analysis of Figs. 5.21 to 5.24, the first conclusion is that the accuracy of the HFTD method is
not affected due to load excitation frequencies close to natural vibration frequency of the system
(resonance condition).
Second, there is a remarkable similarity between the responses of the case studies with excitation
frequencies which are symmetric with respect to the resonance frequency. This is the situation of case
studies A-1 and A-4 on the one hand, and case studies A-2 and A-3 on the other hand, which time
history displacements have similar forms and amplitudes. The only small difference, besides the
obvious fact that the responses of the cases with higher load excitation frequencies have a higher
77

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

number of peaks too, is that the amplitudes of the cases with excitation frequencies below resonance
are slightly greater than its correspondent symmetric cases above resonance.
Finally, it is remarkable the increase of the displacement magnitude for the case studies closer to the
resonance frequency. The maximum displacement magnitude increases from 3 (A-1, A-4) to 8 (A-2, A3), approximately.
Then, based on the good performance shown by the HFTD method in the previous case study, more
critical situations of very lightly damped SDOF linear systems close to resonance condition are
considered. Table 5.25 shows three additional case studies with load excitation frequencies equal to
the resonance frequency and three low damping ratios equal or smaller than 0.05.
Table 5.25: Numerical data of case study B: Resonance excitation frequency with very low
damping ratios
Case
B-1
B-2
B-3

0.05
0.03
0.01

0.85
0.85
0.85

2500
2500
2500

25
25
25

21.19
21.19
21.19

The HFTD parameters for each of the three case studies are summarized in Table 5.26. Similarly than
in case study A, Figs. 5.25 to 5.27 only present the results obtained for displacements, which are
compared with the Direct Integration method response. The analysis of velocity and acceleration
results revealed as good performance as displacement results.
Table 5.26: HFTD parameters for case study B
Case

Tp

B-1
B-2
B-3

0.025
0.025
0.025

42.43
42.43
84.78

1697
1697
3391

NT
0.0236
0.0236
0.0118

1
1
1

Figure 5.25: Resonance case study B-1. Displacement response for Direct Integration and
HFTD solutions.
78

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.26: Resonance case study B-2. Displacement response for Direct Integration and
HFTD solutions.

Figure 5.27: Resonance case study B-3. Displacement response for Direct Integration and
HFTD solutions.

From the analysis of Figs. 5.25 to 5.27, it is clear that the HFTD method remains accurate with very
low damped systems in resonance conditions. In addition, the comparison between Fig. 5.25 ( =
0.05) and Fig. 5.27 ( = 0.01) shows the notorious effect of damping in the amplitude and stability of
motion. The displacement of case study B-3 (system with extremely low damping) growths in time
without bound, until reaching a maximum amplitude equivalent to two and four times the maximum
amplitude of case studies B-2 and B-1, respectively.

79

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

5.3.3 Test examples for SDOF non-linear systems


Three examples of SDOF non-linear systems are solved with the HFTD method (Section 5.1) and
compared with the time domain solution of the Newmark method (Section 5.2.2). The first two
examples present the response of SDOF systems with two different types of hysteretic stiffness nonlinearity. The third test example is equal to the second example with the addition of damping nonlinearity.

5.3.3.1
Linear elasto-plastic non-linear stiffness behavior subject to a half
cycle sine pulse11
The

SDOF

system

described

in

Table

5.27

is

subject

to

the

half

cycle

sine

pulse

shown in Fig. 5.28. The linear elasto-plastic non-linear stiffness force

[
] of the system is shown in Fig. 5.29. Based on the information given in Table 5.27, the
parameters of the SDOF system shown in Table 5.28 are determined.
Table 5.27: Numerical data of Example 5.5 (Chopra, 2007)
k
10

m
0.2533

0.05

Rt
7.5

Rc
-7.5

Table 5.28: Parameters of the SDOF system with stiffness non-linearity subject to sine
pulse loading
Tr
0.90

T1
1.0

6.283

6.275

Figure 5.28: Half cycle sine pulse loading.

11

Example 5.5 of Chopra (2007)

80

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.29: Linear elasto-plastic stiffness non-linearity.

Newmark and HFTD methods are applied in the system using two different time steps in order to
evaluate the improvement of the HFTD accuracy, when reducing the time step size. For the Newmark
method, the Average Acceleration special case was performed, according to the procedure outlined in
Tables 5.4 and 5.5. The parameters required for the execution of Newmark method are shown in
Table 5.29.
Table 5.29: Newmark method (average acceleration) parameters for the SDOF system
with stiffness non-linearity subject to sine pulse loading

t1
0.1

t2
0.01

On the other hand, Table 5.30 shows the values of the parameters required to perform the HFTD
method, outlined in Tables 5.1 and 5.2, for each of the two time step sizes taken into account. In this
case more than one time segment (NT) is required in order to achieve convergence. The number of
time segments required increases for smaller time steps due to the higher amount of time points and
excitation frequencies (N) included in the analysis.
Table 5.30: HFTD parameters for the SDOF system with stiffness non-linearity subject to
sine pulse loading
t

Tp

0.1
0.01

50.00
50.00

500
5000

NT
0.02
0.02

10
13

81

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

The comparison between the Newmark time domain method and the HFTD method solutions for the
displacement, velocity and acceleration is presented in Figs. 5.30 to 5.32. In addition, the non-linear
hysteretic relation between the stiffness force and the displacement is presented in Fig. 5.33.
The most remarkable observation is that all the responses obtained using both methods are almost
identical when a time step (t) equal to 0.01 s is used. On the other hand, when the time step is
equal to 0.1 s, the accuracy of both methods diminishes, being possible to identify small drifts with
respect to the more accurate solution (t = 0.01 s). However this is a shortcoming of the selected
time step size, and not of the particular characteristics of the methods applied. This can be clearly
appreciated in Fig. 5.33, where the non-linear hysteretic elasto-plastic relation between the stiffness
force and the displacement is shown. When an abrupt change of the tangent stiffness is produced, the
proper description of the nonlinear relation is only possible when small time steps are used.

Figure 5.30: Linear elasto-plastic stiffness force non-linearity. Displacement response for
Newmark and HFTD solutions.

Figure 5.31: Linear elasto-plastic stiffness force non-linearity. Velocity response for
Newmark and HFTD solutions.
82

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.32: Linear elasto-plastic stiffness force non-linearity. Acceleration response for
Newmark and HFTD solutions.

Figure 5.33: Linear elasto-plastic stiffness force non-linearity. Hysteretic stiffness forcedisplacement relation for Newmark and HFTD solutions.

83

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

5.3.3.2
Cubic elasto-plastic non-linear stiffness behavior subject to a blast
loading12
The SDOF system described in Table 5.31 is subjected to the blast loading shown in Fig. 5.34.
Table 5.31: Numerical data of Example 8.15 (Humar, 2002)
k
8

m
0.10

0.1118

Rt
8

Rc
-8

Figure 5.34: Blast loading.

The non-linear stiffness force-displacement relation [


] of the system is expressed and shown
in Eq. (5.87) and Fig. 5.35, respectively. Eq. (5.88) shows the initial tangent stiffness equal to 8 which
is obtained through the evaluation of the initial displacement ( ) in the derivative of the forcedisplacement relation.
(5.87)
(5.88)

The nonlinear stiffness diminishes from 8 to 0, which is the point where the stiffness force reaches its
maximum tensile strength equal to 8. After this, the plastic part of the nonlinear relation starts, the
displacement increases and the velocity diminishes until zero. The negative velocity makes the system
going back in the direction of the equilibrium position following a straight line path with a slope equal
to the initial stiffness (see Fig. 5.35). After the stiffness force is reduced to zero, the system enters in
the compression zone under the same cubic elastic relation given by Eq. (5.87). The hysteretic cycle is
repeated until the whole time span of interest is covered.
12

Example 8.15 of Humar (2002).

84

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.35: Cubic elasto-plastic stiffness non-linearity.

Based on the information given in Table 5.31, the parameters of the SDOF system shown in Table
5.32 are determined.
Table 5.32: Parameters of the SDOF system with stiffness non-linearity subject to blast
loading
Tr
0.90

T1
0.702

8.950

8.894

Two different time steps are also used in this test example. For the Newmark method, the Average
Acceleration special case was performed using the parameters shown in Table 5.33. On the other
hand, Table 5.34 shows the values of the parameters required to perform the HFTD method for each
of the two time step sizes taken into account.
Table 5.33: Newmark method (average acceleration) parameters for the SDOF system
with stiffness non-linearity subject to blast loading

t1
0.1

t2
0.01

Table 5.34: HFTD parameters for the SDOF system with stiffness nonlinearity subject to
blast loading
t

Tp

0.1
0.01

35.2
35.13

352
3513

NT
0.0284
0.0285

9
9
85

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

The comparison between the solutions of the Newmark and the HFTD methods for the displacement,
velocity and acceleration responses is presented in Figs. 5.36 to 5.38. In addition, the nonlinear
hysteretic relation between the stiffness force and the displacement is presented in Fig. 5.39.

Figure 5.36: Cubic elasto-plastic stiffness force non-linearity. Displacement response for
Newmark and HFTD solutions.

Figure 5.37: Cubic elasto-plastic stiffness force non-linearity. Velocity response for
Newmark and HFTD solutions.

86

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.38: Cubic elasto-plastic stiffness force non-linearity. Acceleration response for
Newmark and HFTD solutions.

Figure 5.39: Cubic elasto-plastic stiffness force non-linearity. Hysteretic stiffness forcedisplacement relation for Newmark and HFTD solutions.

From the analysis of Figs. 5.36 to 5.39, it is clear that using a time step equal to 0.1 s does not
generate accurate results. The accuracy of the HFTD method is more affected by the time step size
due to the relation that this parameter has with the range of frequencies included in the Fourier
analysis. When the time step is large, the maximum excitation frequency included in the HFTD
analysis is small, and as a consequence, the results obtained are less accurate.
However, similarly than in the previous example, a sufficiently small time step (t = 0.01 s) ensures
the high precision of the results obtained, not only for the HFTD method, but also for the time domain
Newmark method.
87

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Finally, Fig. 5.39 shows that the use of a time step equal to 0.01 s makes it possible to trace the cubic
elastic stiffness force-displacement relation with good precision, which is not possible when using a
larger time step. Furthermore, both Newmark and HFTD methods show that the last two time points
calculated are located in the cubic elastic compression zone. This information is not identified with the
larger time step.

5.3.3.3
Cubic elasto-plastic non-linear stiffness and displacement dependent
non-linear damping behavior subject to a blast loading
The SDOF system is exactly the same of the previous example (Table 5.31) with the only difference
that the damping force is neither linear nor exclusively proportional to velocity, but also displacement
dependent, as expressed in Eq. (5.89). The initial tangent damping is defined by Eq. (5.90), whereas
the value of the constant damping coefficient (c) is calculated with Eq. (5.15). The blast loading used
in this example is also the same as the previous example (Fig. 5.34).

(5.89)

(5.90)

Eq. (5.90) shows that the initial tangent damping can be positive or negative. Both cases are
calculated using Newmark and HFTD methods, but considering just one time step equal to 0.01 s, for
which it has been shown in the previous examples that the more accurate results are obtained. The
values of the parameters c and for the definition of the damping force in Eq. (5.89) are shown in
Table 5.35.
Table 5.35: Parameters for the definition of the nonlinear damping force
c
0.20

On the other hand, the hysteretic nonlinear stiffness force-displacement relation [


] of the
system is the same of the previous example, which is defined by Eq. (5.87) and Fig. 5.35. The
parameters used with the Newmark (average acceleration) method are shown in Table 5.36, whereas
the HFTD parameters are shown in Table 5.37.
Table 5.36: Newmark method (average acceleration) parameters for the SDOF system
with stiffness and damping nonlinearity subject to blast loading

t
0.01

88

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Table 5.37: HFTD parameters for the SDOF system with stiffness and damping
nonlinearity subject to blast loading
t

Tp

0.01

35.13

3513

NT
0.0285

It is important to state that for both cases (negative and positive initial tangent damping) studied in
this example, the parameters of both methods are the same. What is more, the pseudo linear
equation in the frequency domain of the HFTD method (step 1.5.1 of Table 5.2) is always solved using
a positive value of c, no matter the sign of the initial tangent damping given in Eq. (5.90). Therefore,
the difference between one case and the other is concentrated in the expression for calculating the
nonlinear internal damping force in step 1.7 of Table 5.2 (HFTD) and step 2.6 of Table 5.5
(Newmark).
For both cases (positive and negative initial tangent damping), the comparison between the Newmark
time domain method and the HFTD method solutions is presented in Figs. 5.40 to 5.42, for the
displacement, velocity and acceleration, respectively. As in the previous cases, it is shown that using
an adequate time step size (t = 0.01) makes the HFTD method as precise as the Newmark method,
even for this type of highly non-linear systems defined by Eqs. (5.87) and (5.89).
Moreover, the nonlinear hysteretic relations stiffness force-displacement and damping force-velocity
are presented in Figs. 5.43 and 5.44. Both figures confirm that the precision of the HFTD method is
not affected by the type (stiffness or damping non-linearity) and degree (quadratic or cubic) of nonlinearity, neither by the sign of the initial tangent stiffness or damping.

Figure 5.40: Cubic elasto-plastic stiffness force and displacement-dependent damping


non-linearity. Displacement response for Newmark and HFTD solutions.

89

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.41: Cubic elasto-plastic stiffness force and displacement-dependent damping


non-linearity. Velocity response for Newmark and HFTD solutions.

Figure 5.42: Cubic elasto-plastic stiffness force and displacement-dependent damping


non-linearity. Acceleration response for Newmark and HFTD solutions.

90

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.43: Cubic elasto-plastic stiffness force and displacement-dependent damping


non-linearity. Hysteretic stiffness force-displacement relation for Newmark and HFTD
solutions.

Figure 5.44: Cubic elasto-plastic stiffness force and displacement-dependent damping


non-linearity. Hysteretic damping force-velocity relation for Newmark and HFTD
solutions.

5.4 Effect of non-linearity in the response of SDOF systems


The example 8.15 of Humar (2002) was selected to show a comparison in the responses of a SDOF
system with linear elastic, non-linear elasto-plastic stiffness and nonlinear displacement dependent
damping behaviors. The linear response of the system is presented together with the non-linear
91

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

response (previously shown) for the displacement, velocity and acceleration in Figs. 5.45 to 5.47,
respectively. In this section only the HFTD solutions are displayed, taking into account that in the
previous section it was demonstrated that HFTD provides the same results as Newmark method.
In general terms, Fig. 5.45 shows that non-linear behavior increases the magnitude of the
displacement response. The effect of damping non-linearity described by Eq. (5.89) depends on the
sign of the initial tangent damping. If it is positive, the damping force increases since the beginning
and therefore the magnitude of the displacement response is reduced. On the other hand, if the initial
damping is negative, the initial damping force has the same direction of the movement and the
displacement magnitude is amplified.
Figs. 5.46 and 5.47 show that for velocity and acceleration responses, the first positive peak value is
higher for damping non-linearity with negative initial tangent damping. Nevertheless, once the effect
of the initial damping is vanished, the linear elastic behavior provides the higher peaks.

Figure 5.45: Effect of non-linear behavior. Displacement response for HFTD solution.

Figure 5.46: Effect of non-linear behavior. Velocity response for HFTD solution.
92

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

Figure 5.47: Effect of non-linear behavior. Acceleration response for HFTD solution.

93

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

5.

Single Degree of Freedom (SDOF)


Dynamic Systems

94

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

6. SIMPLIFIED ONE-DIMENSIONAL DYNAMIC MODELS FOR DAMRESERVOIR INTERACTION SYSTEMS

6.1. Description of the one-dimensional system


The dam-reservoir interaction system shown in Fig. 6.1 can be modeled in a very simplified way by
just taking into account the one dimensional motion in the horizontal direction.

Figure 6.1: Scheme of a dam-reservoir interaction system.

The structural part of such a one-dimensional model can be represented by a one DOF mass-springdashpot system which reflects its inertial, stiffness and damping properties. On the other hand, the
fluid part can be modeled as a rod which longitudinal motion is described by the one-dimensional
wave equation, the boundary conditions in both ends of the rod, and the physical and geometric
properties of the fluid domain.
This simplified one-dimensional model of a dam-reservoir interaction system excited by a ground
acceleration loading is shown in Fig. 6.2.

Figure 6.2: Simplified one-dimensional model for dam-reservoir interaction systems


95

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

The system shown in Fig. 6.2 is solved using two approaches. The first one is the displacement
approach in the time domain (Section 6.2), whereas the second one is the hydrodynamic pressure in
the frequency domain (Section 6.3), which follows the same formulation explained in Chapter 3 for 3D
MDOF dam-reservoir interaction systems.
In Section 6.4 some dam-reservoir interaction examples are modeled and analyzed with both
approaches. Due to the fact that the accuracy of HFTD method for SDOF systems is proven in Chapter
5, all the simulations developed in Section 6.4 are carried out using this method.

6.2. Displacement approach: non-reflective boundary element


In Fig. 6.2 the fluid part of the system is represented as a one-dimensional continuum medium (rod)
attached to the structure. The movement of the structure interacts with this continuum medium
generating waves that travel longitudinally through all its extension.
One common assumption for the infinite extent boundary condition of the fluid continuum medium is
a radiation or non-reflective condition (See Section 3.1.3.3), which absorb all the energy produced by
the incident waves. This means that at the boundary location no reflective wave traveling in direction
to the structure is generated.
It has been demonstrated that this boundary condition for longitudinal waves can be equivalently
represented by a dashpot (non-reflective boundary element) which viscous damping coefficient (
)
depends on the geometric and physical characteristics of the continuum medium, specifically, on the
rod impedance ( ) (Metrikine and Vrouwenvelder, 2006).
The expression to determine the damping coefficient of the non-reflective boundary element is given
by Eq. (6.1). Therefore, using this approach, the interaction between the fluid and the structure can
be reduced to the effect that the non-reflective boundary condition has in the structural response.
This is achieved by replacing the fluid rod shown in Fig. 6.2 for the dashpot shown in Fig. 6.3, which
damping coefficient is calculated as expressed in Eq. (6.1), in terms of the wave velocity ( ), the fluid
density ( ) and the interaction surface area ( ).
(6.1)

Finally, the HFTD equation of motion of the SDOF model shown in Fig. 6.3 is given by Eq. (6.2) in
terms of the relative displacement amplitude of the dam ( ), the ground displacement amplitude
( ), the pseudo force ( ) containing the systems non-linearity, and the non-reflective viscous
damping (

) due to the interaction with the reservoir1.

] ( )

) ( )

( )

(6.2)

It is important to notice that in Eq. (6.2) the damping force generated by the non-reflective boundary is
proportional to the absolute velocity, whereas the damping force of the structure is proportional only to the
velocity relative to the ground.

96

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

Figure 6.3: Simplified one-dimensional model. Displacement approach with radiation


boundary.

6.3. Pressure approach: Frequency dependent hydrodynamic pressure


The one-dimensional wave equation governing the behavior of the continuum fluid medium shown in
Fig. 6.2 is defined by Eq. (6.3) in terms of the hydrodynamic pressure.

(6.3)

In order to solve this equation and find the analytic expression of the pressure wave propagation, two
boundary conditions (one per each end of the rod) have to be defined. The first boundary condition is
given by the interaction between the structural mass and the fluid, whereas the second one is given
by the hypothesis of the reservoir length (finite or infinite) extension.
The pressure wave equation approach to model the one-dimensional dam-reservoir interaction system
is graphically described in Fig. 6.4. Depending on the assumption for the far extent boundary
condition two different cases can be developed (Sections 6.3.1 and 6.3.2), whereas, if the fluid is
assumed to be incompressible, a third case (Section 6.3.3) is obtained for this approach.

97

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

Figure 6.4: Simplified one-dimensional model. Hydrodynamic pressure approach.

In any of the three particular cases the interaction between the fluid and the structure is reduced to a
frequency dependent pressure or surface load [ ( )] applied to the structural mass-spring-dashpot
system2, as it is shown in Fig. 6.5. For this reason, the one-dimensional model under the pressure
approach is ideal to be solved with the HFTD method.

Figure 6.5: Simplified one-dimensional model. Frequency dependent hydrodynamic


pressure load.

The HFTD equation of motion of the SDOF model shown in Fig. 6.5 is given by Eq. (6.4) in terms of
the relative displacement amplitude of the dam ( ), the ground displacement amplitude ( ), the
pseudo force ( ) containing the systems non-linearity, and the hydrodynamic pressure load ( ) due
to the interaction with the reservoir.

] ( )

( )

( )

( )

(6.4)

This is the same approach used in Chapter 3 for MDOF systems.


98

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

The analytic expressions of ( ) for each of the three particular cases previously mentioned are
derived in the next sections.

6.3.1. Case I: Compressible fluid with radiation boundary of infinite extent


The first boundary condition of Case I is expressed by Eq. (6.5). This equation describes the fluidstructure interaction at the interface (
), being derived from Eq. (3.11) for one-dimensional
motion. Eq. (6.5) denotes the horizontal force equilibrium between the fluid pressure gradient and the
structure inertial force3 per volume.

on

(6.5)

The second boundary condition is expressed by Eq. (6.6) which defines the radiation boundary at
infinite extent (
). This boundary condition, known as the Sommerfelds radiation condition, was
previously presented in Eq. (3.21).
on

(6.6)

The HFTD method solves in the frequency domain the system shown in Fig. 6.4. Consequently, it is
convenient to express the excitation force (ground acceleration) and the response (relative
displacement) of the system in their harmonic form, as shown in Eqs. (6.7) and (6.8), respectively.

( )

(6.7)

( )

(6.8)

In the same way, the pressure wave generated in the fluid domain due to the excitation force in the
structural domain is given by the summation of the incident and reflected harmonic waves. Eq. (6.9)
expresses both harmonic waves in terms of their frequency dependent complex amplitudes and the
wave velocity of propagation.

( )

( )

(6.9)

The expression of the harmonic wave defined in Eq. (6.9) fulfills the wave equation of motion defined
in Eq. (6.3). Therefore, in order to calculate the unknown amplitudes of the incident and reflected
waves, it is required to use both boundary conditions previously defined. After replacing Eq. (6.9) into
the infinite extent radiation boundary condition given by Eq. (6.6), the relations expressed in Eqs.
(6.10) and (6.11) are obtained.

Notice that in Eq. (6.5) the total acceleration is expressed as the sum of ground and relative acceleration.
99

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

( )

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

(6.10)
(6.11)

From Eq. (6.10) it can be concluded that the amplitude of the reflected pressure is equal to zero
[ ( )
], which is expected due to the non-reflective nature of Sommerfelds radiation boundary
condition. Therefore, the pressure wave is only composed by the incident wave, and Eq. (6.9) is
reformulated as expressed in Eq. (6.12).

( )

(6.12)

After replacing Eqs. (6.7), (6.8) and (6.12) into the second boundary condition at
given by Eq.
(6.5), the expression of the incident pressure wave amplitude is obtained in Eq. (6.13). The frequency
dependent amplitude of the hydrodynamic force [ ( )] is obtained in Eq. (6.14) after multiplying the
pressure times the interaction surface area ( ).

( )

[ ( )

( )]

( )

[ ( )

( )]

(6.13)
(6.14)

The HFTD equation for the particular Case I is obtained in Eq. (6.15) by replacing the hydrodynamic
force amplitude given by Eq. (6.14) into Eq. (6.4). It is important to notice that Eq. (6.15) is identical
to Eq. (6.2). This is because that, in spite of being derived by different formulations, both equations
responds to the same hypothesis for the fluid domain, namely, compressible with radiation boundary
of infinite extent.

] ( )

) ( )

( )

(6.15)

Eq. (6.15) shows that for Case I the reservoir interaction with the structure provides additional
damping and frequency dependent loading terms in the left and right-hand sides of the equation,
respectively.

6.3.2. Case II: Compressible fluid with zero pressure boundary of finite extent
The zero pressure boundary condition of finite extent is defined in Eq. (6.16) for a reservoir with finite
length equal to
. After replacing Eq. (6.9) into Eq. (6.16) the relation between the incident and
reflected pressure amplitudes is obtained in Eq. (6.17).
on

(6.16)

100

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

(6.17)

The substitution of Eqs. (6.7) to (6.9) and Eq. (6.17) into the fluid-structure interface boundary
condition, given by Eq. (6.5), provides the expression of the incident pressure amplitude in Eq. (6.18).
The reflected pressure amplitude is obtained in Eq. (6.19) after replacing Eq. (6.17) into Eq. (6.18).

( )

( )

[ ( )

[ ( )

(6.18)

( )]

(6.19)

( )]

The summation of both amplitudes given by Eq. (6.20) provides the total amplitude of the pressure
wave acting at the fluid-structure interface. The frequency dependent hydrodynamic force amplitude
[ ( )] is obtained in Eq. (6.21) after multiplying the pressure times the interaction surface area ( )
and simplify the Euler notation complex numbers division.

( )

( )

)
(

[ ( )

[ ( )

(6.20)

( )]

(6.21)

( )]

The HFTD equation for the particular Case II is obtained in Eq. (6.22) by replacing the hydrodynamic
force amplitude given by Eq. (6.21) into Eq. (6.4). Due to the finite extent reflexive boundary
condition, the fluid interaction with the structure provides exclusively frequency dependent additional
mass to the left and right-hand side of Eq. (6.22).

)
(

] ( )

)
(

) ( )

( )
(6.22)

6.3.3. Case III: Incompressible fluid


For the incompressible fluid case (
), Eq. (6.3) is transformed into Eq. (6.23). This means that
the fluid is not anymore a medium where harmonic pressure waves propagate, but a vibrating body
which amplitudes varies in space. The harmonic form of the hydrodynamic pressure for the
incompressible fluid case is given by Eqs. (6.24) and (6.25).
(6.23)

(6.24)
101

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

(6.25)

Due to the lack of fluids compressibility, radiation boundary condition of infinite extent is not possible
to be considered. Therefore, Eqs. (6.24) and (6.25) are introduced in the boundary condition of finite
extent at
, given by Eq. (6.16). The relation between the constants
and
is obtained in
Eq. (6.26).
(6.26)

Replacing Eqs. (6.7), (6.8), (6.24) and (6.25) into the fluid-structure interface boundary condition
given by Eq. (6.5), and then using the relation of Eq. (6.26), provide the expressions of
and
,
given by Eqs. (6.27) and (6.28), respectively.

[ ( )

(6.27)

( )]

[ ( )

(6.28)

( )]

After replacing Eqs. (6.27) and (6.28) into Eq. (6.25), the space and frequency dependent amplitude
of the hydrodynamic pressure is obtained in Eq. (6.29). However, it is of interest to know what is the
pressure amplitude at
, where the mass of the structure is located. This pressure acting in the
fluid-structure interface [ ( )], as well as its associated force [ ( )], is defined in Eqs. (6.30) and
(6.31), respectively.

(
( )

)[ ( )

(
[ ( )
[ ( )

( )

( )]

(6.29)
(6.30)

( )]

(6.31)

( )]

The HFTD equation for the particular Case III is obtained in Eq. (6.32) by replacing the hydrodynamic
force given by Eq. (6.31) into Eq. (6.4). Eq. (6.32) shows that the hydrodynamic force for the
incompressible fluid case is an additional inertial force provided by the mass of the fluid body (
).
This means that under this assumption, the mass of the fluid behaves like a rigid body oscillating
attached to the structure and introducing no damping to the system, like in the previous particular
cases.

] ( )

) ( )

( )

(6.32)

102

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

6.4. HFTD method numerical examples


A modified version of the HFTD method programmed in MATLAB (The Mathworks, Inc., 2012) 4 for the
solution of the SDOF system examples developed in Chapter 5 is used in this section to obtain the
response of the simplified dam-reservoir interaction system.

6.4.1. Description of the simplified dam-reservoir interaction system


The numerical data describing the characteristics of the one-dimensional dam model shown in Fig. 6.2
is given in Table 6.1. The dams damping coefficient ( ) is calculated with Eq. (6.33), the reservoir
infinite extent non-reflective damping coefficient (
) is calculated with Eq. (6.34), the time that the
incident and reflected pressure wave takes to reach the back of the reservoir and go back to the damreservoir interface (
(

) is calculated with Eq. (6.35), and the mass of water contained in the reservoir

) is calculated in Eq. (6.36).

The ground acceleration loading applied to the base of the dam corresponds to the first 4.5 s of the
El Centro earthquake North-South component shown in Fig. 6.6.
Table 6.1: Numerical data of the one-dimensional dam-reservoir system
Dam
(N/m)
6.6 x 10

0.05

Reservoir
(kg)
3.6 x 107

(m)
2500

(m)
200

(kg/m3)
1000

(m/s)
1440

(6.33)
(6.34)
(6.35)
(6.36)

Figure 6.6: El Centro N-S earthquake ground acceleration.


4

The MATLAB routines can be found in Appendix B.


103

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

6.4.2. Dams linear behavior: Influence of dam-reservoir interaction frequency


dependent properties
The first numerical examples correspond to a linear-elastic behavior of the dam, which is generally the
case for frequency-dependent systems. The information given in Table 6.1 and Eqs. (6.33) to (6.36) is
used by the HFTD method to solve the pseudo-linear equations defined in Eqs. (6.15), (6.22) and
(6.32) for the cases I, II and III, respectively. Additionally, the corresponding SDOF system without
taking into account the hydrodynamic force [ ( )] is also solved as a fourth case which simulates the
analysis of a dam with empty reservoir (dry dam).
Figs. 6.7 to 6.9 show the time history responses for the displacement, velocity and acceleration,
respectively, relative to the base motion. Fig 6.10 presents the variation in time of the base reaction
force, defined as the summation of the dams stiffness and damping forces (
).
Considering that these one-dimensional models constitute a substantial simplification of reality, it is
not possible to obtain general conclusions of the linear behavior of dam-reservoir interaction systems
from Figs. 6.7 to 6.10. However, this information is very illustrating to have an insight of the influence
that frequency dependent properties have in the response of dynamic equations of motion like Eqs.
(6.15), (6.22) and (6.32).
As explained in Section 6.2 and demonstrated in Section 6.3.1, the radiation boundary of infinite
extent (Case I) acts like a dashpot in the system. For that reason, the peak values of all the responses
of Case I are smaller than the corresponding peak values of the dry dam case. For this particular
example, the magnitude of the non-reflective dashpot coefficient [Eq. (6.34)] is so high that one the
earthquake time is finished (4.5 s), the system goes back to the equilibrium position without
oscillations and extremely slow.
The incompressible fluid case (Case III) presents very high peaks for displacement and base reaction
responses. The reason for that is the higher flexibility of the system compared to the dry dam due to
the additional mass provided by the fluid [Eq. (6.32)]. This produces an increment of the length of the
natural vibration period of the system.
One interesting feature of the zero pressure reflexive boundary (case II) is that for all the responses
presented its peak values occur after the earthquake loading has finished. This fact confirms the
reflective characteristic of this boundary condition which does not takes out but introduces energy to
the system. This is the only case which responses do not damp out in time when the loading is not
present anymore.
These observations based on the response of frequency dependent SDOF systems (Figs. 6.7 to 6.10)
provide a general overview of the importance that the reservoir compressibility and boundary
conditions assumptions can have in the seismic analysis of dams.
In the next section the effect that non-linear behavior has in this type of frequency dependent
systems is assessed.

104

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

Figure 6.7: Dam-reservoir interaction SDOF systems. Displacement response for linearelastic behavior.

Figure 6.8: Dam-reservoir interaction SDOF systems. Velocity response for linear-elastic
behavior.

105

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

Figure 6.9: Dam-reservoir interaction SDOF systems. Acceleration response for linearelastic behavior.

Figure 6.10: Dam-reservoir interaction SDOF systems. Base reaction response for linearelastic behavior.

6.4.3. Dams cracking non-linearity: Stiffness softening


The dams concrete cracking non-linearity is represented by the softening cubic relation given by Eq.
(6.37) and schematically shown in Fig. 6.11, between the stiffness force ( ) and the relative
displacement ( ).

(6.37)

106

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

Figure 6.11: Cubic stiffness softening force-displacement relationship for concrete


cracking non-linear behavior

As explained before, Cases I and II are frequency dependent systems which take into account the
reservoir compressibility. Therefore, it is of interest to study the effect of the non-linear behavior in
these two frequency dependent systems. Additionally, the non-linear response of the dry dam case
(no frequency-dependence) is also presented as a reference.
For the three cases the initial tangent stiffness is assumed to be equal to the linear elastic stiffness of
the dam ( ), as expressed in Eq. (6.38). However, the values of the cubic coefficient ( ) in Eq.
(6.37) are assigned for each case as shown in Table 6.2.
(6.38)

( )

Table 6.2: Values assigned to the cubic coefficient


Dry dam
2.40 x 1012

of Eq. (6.37) for each analysis case

Compressible water with radiation


boundary of infinite extent
(Case I)
2.40 x 1012

Compressible water with


reflective boundary of finite
extent (Case II)
1.05 x 1012

The previous analysis of Figs. 6.7 to 6.10 demonstrates that the reflective boundary condition (Case
II) generates higher linear-elastic responses, and therefore a stiffer and more resistant non-linear
behavior is required for this particular system. This is the reason why a smaller value of
is assigned
to this case in Table 6.2.
Finally, the pseudo force in the time domain for the cubic stiffness softening non-linearity is calculated
using Eq. (6.39). Then, the non-linear behavior is completely defined and the HFTD method can be
applied in the solution of the SDOF dam-reservoir interaction models.

( )

(6.39)

107

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

First, the nonlinear behavior of the dry dam case is studied as a reference of a non frequency
dependent system. Figs. 6.12 and 6.13 present the effect of non-linear behavior in the displacement
and stiffness internal force of the SDOF system, respectively.

Figure 6.12: Dry dam case. Effect of non-linearity in the displacement response.

Figure 6.13: Dry dam case. Effect of non-linearity in the stiffness force response.

Fig 6.12 shows that the major effects of non-linear behavior occur between 2 s and 4 s. As expected,
the system including softening non-linearity generates the higher displacement peak value after the
time 2 s. Fig 6.13 shows that this is the most far-away point located in the plastic regime (horizontal
branch), where the displacement of the system increases with no additional resisting stiffness force.
After the time 4.5 s the ground acceleration loading ends and the system comes back in free vibration
to the linear elastic regime, where both linear and non-linear responses damp-out with a small phase
and the same magnitude.

108

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

Second, the nonlinear behavior of the frequency dependent Case I, representing a dam-reservoir
interaction system with a radiation boundary of infinite extent, is studied. Figs. 6.14 and 6.15 present
the effect of non-linear behavior in the displacement and stiffness internal force, respectively, of this
SDOF frequency dependent system.

Figure 6.14: Dam-reservoir interaction with infinite extent radiation boundary (Case I).
Effect of non-linearity in the displacement response.

Figure 6.15: Dam-reservoir interaction with infinite extent radiation boundary (Case I).
Effect of non-linearity in the stiffness force response.

Due to the additional damping introduced to the system by the infinite extent boundary condition [Eq.
(6.15)], the influence of non-linear behavior is less important in this case than in the dry dam case.
Certainly, Fig. 6.14 shows that there are increments of the displacement response peak values around
time 2 s, however these are not as significant as in Fig. 6.12. The same can be concluded from Fig.
109

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

6.15 which shows that the displacements produced are not large enough to reach the horizontal
yielding branch and develop a more important non-linear behavior.
Finally, the nonlinear behavior of the frequency dependent Case II, representing a dam-reservoir
interaction system with a reflective boundary of finite extent, is studied in Figs. 6.16 and 6.17 which
present the effect of non-linear behavior in the displacement and stiffness internal force, respectively.

Figure 6.16: Dam-reservoir interaction with finite extent reflective boundary (Case II).
Effect of non-linearity in the displacement response.

Figure 6.17: Dam-reservoir interaction with finite extent reflective boundary (Case II).
Effect of non-linearity in the stiffness force response.

In this case the non-linear behavior is more important than in the radiation boundary case. As
explained before, the reflective boundary condition introduces energy in the system generating large
displacements even after the loading duration. Fig. 6.16 shows that all the peak values of the
displacement are considerable higher for the non-linear behavior. It is important to notice that the
110

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

yielding branch of the stiffness force is reached only after time 5 s where the ground acceleration
loading is not present anymore.
One final observation valid for the three cases presented is that for this type of loading the non-linear
behavior of the system influences only the magnitude but not the profile of the time history response.
Figs. 6.12, 6.14 and 6.16 prove that this is true no matter if the system is frequency dependent or
not.

111

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

6. Simplified One-Dimensional Dynamic


Models for Dam-Reservoir Interaction Systems

112

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

7. IMPLEMENTATION OF THE HFTD METHOD IN DIANA FOR THE


SEISMIC ANALYSIS OF FOUNDATION-DAM-RESERVOIR
INTERACTION PROBLEMS

7.1. Definition of foundation-dam-reservoir interaction models in DIANA 1

7.1.1. Flow finite elements


Flow finite elements are used in DIANA for solving all types of convection-diffusion problems. For this
particular case, general flow elements are used to model the water of the reservoir.
The basic variable of a general flow element are scalar potentials in their nodes, which for fluidstructure interaction models are pressure potentials (p). The derivative of these pressure potentials in
each direction of analysis produces fluxes (qx, qy, qz) which are located in the integration points.
Different types of 2D, 3D and axisymmetric flow elements can be defined depending on the
characteristics of the model. For each type of fluid element a particular boundary element should be
assigned to the fluid domain in order to define the boundary conditions.
Quadrilateral flow elements Q4HT and CQ8HT shown in Fig. 7.1 are presented and used in the
modeling of two-dimensional dam-reservoir interaction modeling. The pressure field inside the flow
presents a linear variation for Q4HT and quadratic for CQ8HT.

Figure 7.1: DIANA 2D flow finite elements.2

For each type of 2D flow element a specific flow boundary element should be attached in the edges of
the fluid domain. A particular flux (or pressure derivative) in the direction perpendicular to the
reservoir edge is assigned to the flow boundary element, depending on the type of boundary
condition required to be modeled. As explained in Chapter 3, for the dam-reservoir interaction
problem these boundary conditions are bottom absorption, radiation boundary of infinite extent and
surface waves (See Section 3.1.3).
1
2

Based on TNO DIANA (2011).


Figures taken from TNO DIANA (2011).
113

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Straight and curved 2D flow boundary elements, B2HT and BC3HT, respectively, are shown in Fig.
7.2. The flux qy perpendicular to the boundary presents a linear variation along its length for B2HT
and a quadratic variation for BC3HT.

Figure 7.2: DIANA 2D boundary flow finite elements.3

Similarly than for the two-dimensional models, Fig. 7.3 shows two types of flow finite elements for
three-dimensional models, namely HX8HT and CHX20H, respectively. BQ4HT and BCQ8HT are their
corresponding flow boundary elements which are shown in Fig. 7.4.

Figure 7.3: DIANA 3D flow finite elements.4

3
4

Figures taken from TNO DIANA (2011).


Figures taken from TNO DIANA (2011).
114

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Figure 7.4: DIANA 3D boundary flow finite elements.5

Two material parameters should be defined for the flow elements in a fluid-structure interaction
model. The first one is the conductivity (cF) of the fluid which in this type of analysis is a dummy
parameter which definition allows the calculation of the fluid domain conductivity matrix
. If in
addition the compressible effects are desired to be taken into account, then the sonic speed (c) should
be defined for the compressibility matrix
to be implemented (See Section 3.1.5).

7.1.2. Fluid-structure interface


Fluid-structure interface elements are located in the common contact surface (or line) between the
fluid and the structure domain. This element couples both domains by ensuring the equilibrium and
continuity between the fluid pressure and the solid displacements normal to the interface.
In DIANA there are two types of fluid-structure interface elements: 1) Line interface elements for 2D
or axisymmetric models; 2) Plane interface elements for 3D models. Schematic examples of both
types of fluid-structure interface connecting structural and flow elements are shown in Fig. 7.5.

Figure 7.5: DIANA fluid-structure interface elements.6


5
6

Figures taken from TNO DIANA (2011).


Figures taken from TNO DIANA (2011).
115

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

One side of the fluid-structure interface elements are defined in terms of fluid pressure variables in a
direction perpendicular to the interaction surface, whereas the other side is defined in terms of
structure displacements variables in the three (or two) dimensions of analysis. For example, the
hybrid variables definition of the fluid-structure interface elements is shown in Fig. 7.6 for the line
interface BL4S2 and the plane interface BQ12S4.

Figure 7.6: DIANA fluid-structure interface elements variables.7

One material parameter is required by DIANA for the definition of a fluid-structure interface property,
which is the fluid density ( ).

7.1.3. Reservoir boundary conditions


Specific material parameters should be assigned to the DIANA boundary flow elements (Section 7.1.1)
in order to model the different types of reservoir boundary conditions explained in detail in Section
3.1.3.
Two types of boundary conditions were used in the present work: bottom absorption boundary and
radiation boundary of infinite extent. The two material parameters required to be assigned to the
boundary flow elements in order to model these two types of boundary conditions are the sonic speed
(c) and the wave reflection coefficient ( ), which value varies between -1 (fully absorptive) and 1
(fully reflective). For the infinite extent radiation boundary, the wave reflection coefficient is always
equal to zero (
).

Figures taken from TNO DIANA (2011).


116

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

The definition of these parameters is used in the calculation of the radiation matrix
3.1.5).

(See Section

7.1.4. Foundation modeling


The soil which acts as the dam foundation is modeled with the same type of structural finite elements
used to model the dam. Generally, plain strain elements are suitable for 2D models, whereas solid
elements for 3D models.
In DIANA the foundation elements can be directly connected to the base of the dam if no base sliding
analysis is considered. The flexibility of the foundation is taken into account by defining its linear
elastic behavior parameters (f.e.: Young modulus, poison ratio), which values depend on the type of
foundation (f.e.: sand, clay, rock).
The mass of the foundation is not necessarily taken into account. In structural dynamics the massless
foundation is a well-established approach in which the stiffness of the foundation is included in the
analysis but keeping out its mass. In this way, the foundation only plays the role of a boundary
condition for the structure, and not a vibrating mass inside the model.
In DIANA the radiation far field boundary of the soil can be modeled using boundary surface
elements. This is a special element to model the free field medium in dynamic analysis 8. Another way
to do this is by means of viscous boundaries consisting on concentrated dashpots located in the
foundation edges. The damping coefficients are calculated for both directions, normal and tangential
to the boundary surface.
In the present work the radiation far field boundary of the foundation is not taken into account. In
turn the size of the soil is considerably extended in such a way that the soil boundary reflective effects
have no influence in the dams vibration.

7.1.5. Dams concrete cracking non-linear behavior


The total strain rotating crack model was used in DIANA to model the non-linear concrete cracking
behavior of the dam. Under this approach both the tensile and compressive behavior of the concrete
is defined by means of a stress-strain curve which is used for determining the non-linear internal force
vector, and therefore the pseudo force vector required by the HFTD method (See Chapter 4), for each
time step.
For the concrete tensile behavior different types of predefined tension softening functions shown in
Fig. 7.7 can be chosen in DIANA. Similarly, for the concrete crushing behavior the predefined
compression functions shown in Fig. 7.8 are available. Depending on the type of function, different
parameters are required to be defined9.

8
9

For detailed information about boundary surface elements see TNO DIANA (2011).
For detailed information about total strain crack models see TNO DIANA (2011).
117

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Figure 7.7: DIANA predefined tension softening functions for total strain crack model.10

Figure 7.8: DIANA predefined compression functions for total strain crack model. 11

7.2. HFTD implementation in DIANA


Hybrid Frequency-Time Domain method was programmed, codified and implement in DIANA
based on the theoretical background of Chapter 4. The flowcharts of the HFTD implementation in
DIANA are reproduced in Appendix A. In addition, the command file for the execution of the HFTD
analysis in DIANA is shown in Appendix C.

10
11

Figures taken from TNO DIANA (2011).


Figures taken from TNO DIANA (2011).
118

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

7.3. Case study: 2D foundation-dam-reservoir interaction model

7.3.1. Case study description


The finite element mesh of the 2D case study is shown in Fig. 7.9. The model is composed of three
parts which are the dam (cyan), the reservoir (yellow) and the foundation (green). The gravity dams
height is 100 m, whereas its base length is 80 m. The reservoirs length is considered to be 390 m,
whereas the length and depth of the soil foundation block are 720 m and 200 m, respectively.

Figure 7.9: DIANA model of the 2D foundation-dam-reservoir interaction case study.

The types of finite elements used for each component of the model are indicated in Table 7.1. The
dam and foundation elements are the typical quadrilateral (Q8EPS) and triangular (T6EPS) plane
strain elements with linear interpolation. The characteristics of the reservoir flow and fluid-structure
interface elements are explained in Section 7.1.
The values of the linear and non-linear material parameters used in this case study for the dam, the
foundation and the reservoir (fluid domain and boundary parameters) are listed in Tables 7.2, 7.3 and
7.4, respectively.
Table 7.1: Type of DIANA finite elements used for the case study modeling
Component

Type

Element

Dam

Plane strain 2D

Q8EPS
T6EPS

Foundation

Plane strain 2D

Q8EPS
T6EPS

Reservoir

General flow 2D

Q4HT

Reservoir boundaries

Flow boundary 2D

B2HT

Reservoir-dam interface

Fluid-structure interface 1D

BL4S2
119

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Table 7.2: Material parameters for the concrete dam


DIANA variable name

Value / Type

Units

Modulus of elasticity

YOUNG

2.7 x 1010

N / m2

Poison modulus

POISON

1.67 x 10-1

Density

DENSIT

2.4 x 103

kg / m3

Rayleigh damping

RAYLEI

8.19456 x 10-1

1.38396 x 10-3

Softening tension curve

TENCRV

LINEAR

Tensile strength

TENSTR

3.8 x 106

N / m2

Ultimate tensile strain

EPSULT

5.0 x 10-4

Compression curve

COMCRV

CONSTA

Compressive strength

COMSTR

3.5 x 107

N / m2

DIANA variable name

Value / Type

Units

Modulus of elasticity

YOUNG

2.5 x 1010

N / m2

Poison modulus

POISON

2.0 x 10-1

Density

DENSIT

kg / m3

DIANA variable name

Value / Type

Units

Conductivity

CONDUC

2.5 x 1010

N / m2

Sonic speed

CSOUND

1.44 x 103

m/s

Density

DENSIT

1.0 x 103

kg / m3

Wave reflection coefficient for


infinite extent boundary

ALPHAB

Wave reflection coefficient for


bottom absorption boundary

ALPHAB

5.0 x 10-1

Parameter

Table 7.3: Material parameters for the foundation soil


Parameter

Table 7.4: Material parameters for the reservoir fluid


Parameter

Finally, the foundation-dam-reservoir interaction system is subject to a base acceleration loading in


the horizontal direction which is applied in the supports located in the foundation edges (See Fig. 7.9).
120

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

For transient and HFTD analysis the El Centro earthquake ground acceleration signal shown in Fig.
7.10 is defined as the loading input. On the other hand, for direct frequency analysis the
corresponding ground acceleration spectrum shown in Fig. 7.11 is taken into account.

Figure 7.10: El Centro N-S earthquake ground acceleration signal.

Figure 7.11: El Centro N-S earthquake ground acceleration spectrum.

7.3.2. Eigen-analysis
The eigen-analysis of the dam was performed following the two procedures presented and explained
in Section 3.1.9. The first procedure, known as dry dam eigen-analysis, does not take into account
the influence of the reservoir in the calculations, and therefore, it solves the free vibration equation of
motion of the structures given by Eq. (3.73) in terms of the mass ( ) and stiffness ( ) matrices.

(3.73)
121

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

On the other hand, the second procedure, known as wet dam eigen-analysis, does take into account
the influence of the fluid reservoir but assuming that it is incompressible and therefore, without
radiation but reflective boundary conditions. In this approach the reservoir introduces in the free
vibration equation of motion, given by Eq. (3.75), a fluid added mass term ( ) defined by Eq. (3.72)
(See Chapter 3).

(3.72)

(3.75)

The dry dam eigen-frequencies and effective mass participation in both directions of analysis are
presented in Fig. 7.12 for the first 15 modes of vibration. The fundamental vibration frequency for the
dam without reservoir interaction is 3.33 Hz. Besides, it can be noticed that the first modes of
vibration concentrate more than 90% of the vibration mass in both directions, being the first and third
modes the most important for X and Y direction, respectively.

Figure 7.12: Eigen-analysis results for the empty reservoir or dry dam case

The total mass of the dam is equal to 9.6 x 106 kg. When the wet dam eigen-analysis is performed,
the calculated fluid added mass is equal to 5.42 x 10 6, which means a mass increment of more than

122

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

50%. This considerable additional mass makes the system more flexible and therefore the eigenfrequencies obtained are smaller than in the dry dam case.
Indeed, this is confirmed by the fundamental eigen-frequency for the wet dam case shown in Fig.
7.13, which is reduced to 2.62 Hz. However, even since the first mode, the effective mass
participating in the X-direction is higher than 115%, which means that the fluid added mass actively
contributes in the dams horizontal vibration. This is not the case for the vertical vibration (Y-direction)
where the fluid mass has no participation at all, due to the vertical orientation of the dam-reservoir
interface.

Figure 7.13: Eigen-analysis results for the full reservoir or wet dam case

The remarkable influence of the fluid added mass in the eigen-frequency values is not replicated in
the mode shapes. Fig. 7.14 shows the modes shapes of the 1st, 2nd and 4th vibration modes in the Xdirection. It can be noticed that the first mode shape is not influenced by the added fluid mass, no
matter that the eigen-frequency experiences a considerable reduction. Small magnitude differences
can be appreciated in the second and fourth mode shapes.
It is due to these negligible differences that the HFTD method uses the mode shapes of the empty
reservoir eigen-analysis for the mode-superposition method.

123

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Figure 7.14: First three mode shapes in the X-direction for empty (eigen-dry) and full
(eigen-wet) reservoir cases

Finally, the influence of the contribution of the initial conditions in the geometric stiffness matrix, and
therefore in the eigen-analysis results, was investigated. For this purpose, the gravity load of the dam
and the hydrostatic pressure applied in the dams face and in the soil below the reservoir were
included as initial conditions.

124

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

As shown in Figs. 7.15 and 7.16 for the dry and wet dam cases, respectively, the influence of the
initial loading conditions in the stiffness matrix of the structure is not significant. For this reason, the
eigen-analysis results presented in Figs. 7.15 and 7.16 are almost the same as those presented in
Figs. 7.12 and 7.13.

Figure 7.15: Eigen-analysis results for the empty reservoir or dry dam case including
initial loading conditions

125

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Figure 7.16: Eigen-analysis results for the full reservoir or wet dam case including initial
loading conditions

7.3.3. Frequency domain linear elastic analysis


With the objective of evaluate the effect of the frequency dependent properties introduced to the
dam-reservoir interaction system by fluid compressibility and reservoir boundary conditions, like
infinite radiation extent and bottom absorption, a linear elastic frequency domain analysis using the
Direct Method12 is performed for the following five cases which are illustrated in Fig. 7.17.

Case
Case
Case
Case
Case

I: Incompressible fluid.
II: Compressible fluid with reflective reservoir boundaries.
III: Compressible fluid with bottom absorption boundary.
IV: Compressible fluid with infinite extent radiation boundary.
V: Compressible fluid with bottom absorption and infinite extent radiation boundaries.

The ground acceleration spectrum shown in Fig. 7.11 is applied to the five systems shown Fig. 7.17,
and as a consequence, the Amplitude-Frequency response plots given by Figs. 7.18 to 7.21 give an
12

For theoretical background of the frequency domain analysis see Chapter 2.


126

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

illustrative insight of the effect that compressibility and reservoir boundary conditions have in the
dynamic behavior of dam-reservoir interaction systems.

Figure 7.17: Direct frequency linear elastic analysis cases.

Figs. 18 and 19 show the relative Displacement Amplitude-Frequency plot, in the X and Y directions
respectively, of the dams crest node (higher point of the dam).

Displacement Amplitude X-direction


(m)

0.014

C1_incomp

0.012

C2_compres
C3_compres_botabs

0.01

C4_compres_infext

0.008

C5_compres_botabs_infext
0.006
0.004
0.002
0
0

10

11

Frequency (1/s)
Figure 7.18: X-direction Displacement Amplitude-Frequency plot of the dams crest.
127

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Displacement Amplitude Y-direction


(m)

0.0045

C1_incomp

0.004

C2_compres

0.0035

C3_compres_botabs

0.003

C4_compres_infext

0.0025

C5_compres_botabs_infext

0.002
0.0015
0.001
0.0005
0
0

10

11

Frequency (1/s)
Figure 7.19: Y-direction Displacement Amplitude-Frequency plot of the dams crest.

For all the case studies the higher amplitudes are obtained around the fundamental vibration
frequency of the wet-dam system equal to 2.62 Hz (See Figs. 7.13 and 7.16).
Comparing the cases C1 and C2 it can be concluded that assuming that the fluid of the reservoir is
compressible produces an increment of the displacement amplitudes. The opposite occurs in case C3
for which a bottom absorption boundary condition is introduced. This is due to the additional damping
introduced to the system by the bottom absorption (See Chapter 3).
On the other hand, in this case study the introduction of infinite extent radiation boundary condition
(cases C4 and C5) does not influences decisively the magnitude of the responses around the
fundamental frequency of vibration. Only for the higher vibration frequencies around 6 Hz the
introduction of the infinite extent radiation boundary damps out the peaks of Cases C1 and C2. The
main reason for the negligible contribution of the infinite extent boundary in this case study is the
long extension of the reservoir, defined in the model and equal to 390 m.
Finally, Figs. 20 and 21 present the relative velocity and acceleration, respectively, of the dams crest
node in the X-direction. The same observations made for the displacement response are applicable to
both velocity and acceleration amplitudes, with the only difference that the importance of
compressibility and infinite extent radiation boundary is considerably higher for acceleration (Fig. 21)
than for any other response.

128

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Velocity Amplitude X-direction (m/s)

0.25

C1_incomp
C2_compres

0.2

C3_compres_botabs
C4_compres_infext

0.15

C5_compres_botabs_infext
0.1

0.05

0
0

10

11

Frequency (1/s)
Figure 7.20: X-direction Velocity Amplitude-Frequency plot of the dams crest.

3.5

C1_incomp

Acceleration Amplitude X-direction


(m/s2)

C2_compres
C3_compres_botabs

2.5

C4_compres_infext
2

C5_compres_botabs_infext

1.5
1
0.5
0
0

10

11

Frequency (1/s)
Figure 7.21: X-direction Acceleration Amplitude-Frequency plot of the dams crest.

7.3.4. Transient analysis: Linear and non-linear seismic behavior


Transient linear and non-linear analysis of the 2D foundation-dam-reservoir interaction system is
performed for the following three cases which are described in Fig. 7.22.

Case I: Empty reservoir (influence fluid-structure interaction not considered).


Case II: Incompressible fluid (no frequency dependent properties)
Case III: Compressible fluid with reflective reservoir boundaries.
129

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Figure 7.22: Transient linear and non-linear analysis cases.

The ground acceleration loading shown in Fig. 7.10 is applied to the three case studies and their
responses are determined using the Newmark (average acceleration) and HFTD methods.
The type of cracking non-linearity, considered exclusively for the dams concrete, is described in
Section 7.1.5, whereas the values of the concrete non-linear parameters are presented in Table 7.2.

7.3.4.1.

Case I: Empty reservoir

A comparison of the structural concrete cracking non-linear responses obtained in the dams crest
employing both Newmark and HFTD methods is presented in Figs. 7.23 to 7.25 for the relative
displacement, velocity and acceleration responses in the X-direction, respectively.
The empty reservoir is an ideal case in which the influence of the dynamic interaction of the reservoir
with the dam is neglected. For this reason, the magnitudes of the responses obtained are small and
the system barely reaches the non-linear behavior. Actually, most of the behavior shown in Figs. 7.23
to 7.25 corresponds to a linear elastic behavior.
The HFTD and Newmark results coincide very well until the time 2.5 s. After this time step, the slight
cracking of concrete begins and the difference between both responses starts to be noticeable. The
reason for this difference other is that, in spite of being small in both methods, the extension and
magnitude of concrete cracking in Newmark is greater than in HFTD.
Fig. 7.26 shows the cracking pattern of HFTD and Newmark methods for the last time step of the
simulation equal to 4.49 s. It can be noticed that as expected the small cracking zone is located for
both methods in the foot of the dam in the side of the reservoir. However, Newmark solution develops
a more extended cracking zone for which the cracking strains are also higher in comparison with the
130

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

HFTD solution. An explanation for the inherent different non-linear performance of both methods is
developed in the full reservoir with incompressible fluid case (Section 7.3.4.2).

Displacement X-direction (m)

0.05

Newmark_nlcrack

0.04

HFTD_nlcrack

0.03
0.02
0.01
0
-0.01
-0.02
-0.03
-0.04
0

0.5

1.5

2.5

3.5

4.5

Time (s)
Figure 7.23: Empty reservoir concrete cracking non-linear response. Dams crest
displacement X-direction.

0.8

Newmark_nlcrack

Velocity X direction (m/s)

0.6

HFTD_nlcrack

0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
-1.2
0

0.5

1.5

2.5

3.5

4.5

Time (s)
Figure 7.24: Empty reservoir concrete cracking non-linear response. Dams crest velocity
X-direction.

131

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Acceleration X direction (m/s2)

30

Newmark_nlcrack
20

HFTD_nlcrack

10
0
-10
-20
-30
0

0.5

1.5

2.5

3.5

4.5

Time (s)
Figure 7.25: Empty reservoir concrete cracking non-linear response. Dams crest
acceleration X-direction.

Figure 7.26: Empty reservoir concrete cracking non-linear response. Cracking pattern at
time 4.49 s.

7.3.4.2.

Case II: Full reservoir with incompressible fluid

When the reservoir interaction assuming incompressible fluid is considered, the responses of the dam
become greater13. First a comparison of the structural linear elastic responses obtained in the dams

13

The influence of the reservoir interaction in the response of the dam was analytically presented for the onedimensional models of Chapter 6.
132

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

crest employing both Newmark and HFTD methods is presented in Figs. 7.27 to 7.29 for the relative
displacement, velocity and acceleration in the X-direction, respectively.

Displacement X-direction (m)

0.1

Newmark_linelast

0.08

HFTD_linelast

0.06
0.04
0.02
0
-0.02
-0.04
-0.06
-0.08
0

0.5

1.5

2.5

3.5

4.5

5.5

Time (s)
Figure 7.27: Incompressible fluid linear elastic response. Dams crest displacement Xdirection.

1.5

Velocity X direction (m/s)

Newmark_linelast
1

HFTD_linelast

0.5
0
-0.5
-1
-1.5
0

0.5

1.5

2.5

3.5

4.5

5.5

Time (s)
Figure 7.28: Incompressible fluid linear elastic response. Dams crest velocity X-direction.

133

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Acceleration X direction (m/s2)

40

Newmark_linelast

30

HFTD_linelast

20
10
0
-10
-20
-30
-40
0

0.5

1.5

2.5

3.5

4.5

5.5

Time (s)
Figure 7.29: Incompressible fluid linear elastic response. Dams crest acceleration Xdirection.

Figs. 7.27 to 7.29 show a very good agreement in the linear elastic solutions obtained with both
methods. However, small differences are appreciated in the peak values since time 3.5 s thereafter,
especially for the acceleration response.
As it is explained in the following lines for the non-linear response, the reason for these differences is
related with the choice of implementing the HFTD method in combination with the mode superposition
method. Not all the modes of vibration but only fifteen are included in the HFTD analysis, and
therefore part of the solution contributing with the peaks response is excluded.
Furthermore, the loading input signal shown in Fig. 7.10 has a maximum sampling frequency equal to
25 Hz. This means that any excitation frequency contained in the loading, or any modal vibration
frequency of the system, which is greater to 25 Hz is not included in the HFTD response.
Before showing the results obtained for the concrete cracking non-linear behavior of the dam, the
hydrodynamic pressures generated by the linear elastic dam-reservoir interaction is shown for
different time steps in Fig. 7.30. It is clear that the hydrodynamic pressures can assume several types
of height distribution depending of the acceleration in the interface nodes for each time step of the
analysis.

134

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Figure 7.30: Incompressible fluid linear elastic response. Hydrodynamic pressures at


different time steps.

Now, a comparison of the structural concrete cracking non-linear responses obtained in the dams
crest, employing both Newmark and HFTD methods, is presented in Figs. 7.31 to 7.33 for the relative
displacement, velocity and acceleration in the X-direction, respectively.
As mentioned before, the consideration of a full reservoir with incompressible fluid in the seismic
analysis produces a significant increment of the responses amplitudes. For example, Fig. 7.31 shows
that the maximum displacement obtained in this case study is close to between 0.06 m and 0.08 m.
On the other hand, Fig. 7.23 shows that when the reservoir interaction with the dam is neglected, the
maximum non-linear displacement is less than 0.04 m. This means that the level of cracking is also
intensified when the incompressible fluid is taken into account, and as a consequence the difference
between the cracking non-linear behaviors predicted by HFTD and Newmark methods is more obvious
than in the empty reservoir case.

135

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Displacement X-direction (m)

0.1

Newmark_nlcrack

0.08

HFTD_nlcrack

0.06
0.04
0.02
0
-0.02
-0.04
-0.06
-0.08
0

0.5

1.5

2.5

3.5

4.5

5.5

Time (s)
Figure 7.31: Incompressible fluid concrete cracking non-linear response. Dams crest
displacement X-direction.

1.5

Velocity X direction (m/s)

Newmark_nlcrack
1

HFTD_nlcrack

0.5
0
-0.5
-1
-1.5
0

0.5

1.5

2.5

3.5

4.5

5.5

Time (s)
Figure 7.32: Incompressible fluid concrete cracking non-linear response. Dams crest
velocity X-direction.

136

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Acceleration X direction (m/s2)

40

Newmark_nlcrack

30

HFTD_nlcrack

20
10
0
-10
-20
-30
-40
0

0.5

1.5

2.5

3.5

4.5

5.5

Time (s)
Figure 7.33: Incompressible fluid concrete cracking non-linear response. Dams crest
acceleration X-direction.

Figs. 7.31 to 7.33 indicate that the non-linear behavior described by the HFTD method is moderate,
resembling what is a known as a mild type of non-linearity. For this reason, the shape of the timehistory plots is similar to the ones of the linear elastic behavior (Figs. 7.27 to 7.29), but with slightly
smaller amplitudes.
On the other hand, the non-linear behavior described by the Newmark method is more severe,
diminishing considerably the response amplitudes and altering the periods of oscillations. The
substantial difference between the non-linear cracking behaviors predicted by each method at
different time steps is illustrated in Fig. 7.34.
It can be noticed that in the time step 1.42 s, where the first crack appears, there is a good
agreement between both methods about which is the zone with higher tensile stresses. However,
during the following time steps, and especially around 2.5 s, the Newmark method develops a very
pronounced crack in the base due to the concentration of high tensile strains nearly that zone.
In spite of calculating a similar displacement field as Newmark does, HFTD method does not predict
such a tensile strain concentration in the base of the dam. Therefore, the HFTD tensile strain field is
uniformly distributed through the entire dams body and has considerably smaller magnitudes.
This observation about the displacement and strain fields obtained with both methods is graphically
demonstrated in Fig. 7.35 for the time step 2.40 s, when the big cracks shown in Fig. 7.34 for the
time step 2.61 s have not appeared yet.

137

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Figure 7.34: Incompressible fluid reservoir non-linear response. Dams concrete cracking
pattern at different time steps.

The large crack in the base of the dam calculated by the Newmark method (Fig. 7.34) is the reason
why the responses shown in Figs. 7.31 to 7.33 are highly non-linear for this method. What is more,
due to the position and magnitude of the concrete cracking, the dam acts like an isolated system in
which the ground acceleration loading is not fully transmitted to the structure. For this reason, since
the time step 2.5 s the magnitudes of the responses in the crest of the dam calculated with the
Newmark method diminish substantially.
On the other hand, due to the fact that in the HFTD method the non-linear response of the dam is
calculated based on the superposition of the linear elastic modes of vibration, the higher tensile
strains are predicted to be located in the vertical face of the dam interacting with the reservoir and
not in the base. In this case, the cracking is more distributed and therefore less severe. Furthermore,

138

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

the location of the higher tensile zones seems to be ruled by the first and the third mode shapes of
vibration in the horizontal direction (See Fig. 7.14).

Figure 7.35: Incompressible fluid reservoir non-linear response. Displacement and


principal strain fields at time step 2.40 s.

7.3.4.3.

Case III: Full reservoir with compressible fluid

The frequency dependent dam-reservoir interaction system with compressible fluid and infinite extent
radiation boundary condition (Case III in Fig. 7.22) is directly solved in DIANA only by the Direct
Method (linear frequency domain solution) and the HFTD method (non-linear transient solution). Due
to the frequency dependence of the system, Newmark method is not applicable for this case.
Therefore, non-linear concrete cracking HFTD solution is presented in Figs. 7.36 to 7.38 for the dams
crest relative displacement, velocity and acceleration, respectively, in the X-direction. A general
observation of the time history graphics allows concluding that for this particular model the inclusion
of compressibility in combination with the radiation boundary of infinite extent does not change
substantially the peak values of the non-linear responses of the system. This analyzed in detail in the
next section.
139

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Displacement X direction (m)

0.1

HFTD_nlcrack

0.08
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
-0.08
0

0.5

1.5

2.5

3.5

4.5

Time (s)
Figure 7.36: Compressible fluid with radiation boundary. HFTD non-linear solution for the
dams crest displacement in X-direction.

1.5

Velocity X direction (m/s)

HFTD_nlcrack
1
0.5
0
-0.5
-1
-1.5
0

0.5

1.5

2.5

3.5

4.5

Time (s)
Figure 7.37: Compressible fluid with radiation boundary. HFTD non-linear solution for the
dams crest velocity in X-direction.

140

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Acceleration X direction (m/s2)

30

HFTD_nlcrack
20
10
0
-10
-20
-30
0

0.5

1.5

2.5

3.5

4.5

Time (s)
Figure 7.38: Compressible fluid with radiation boundary. HFTD non-linear solution for the
dams crest acceleration in X-direction.

7.3.4.4.

HFTD results analysis: Influence of the dam-reservoir interaction


in the non-linear structural response

The influence of dam-reservoir interaction and fluid compressibility in the dams concrete cracking
non-linear response is analyzed based on the HFTD solutions of the three cases previously studied
(Fig. 7.22). The dams crest relative displacement, velocity and acceleration in the X-direction are
presented in Figs. 7.39 to 7.41, respectively.

Displacement X-direction (m)

0.1

C1_Empty reservoir

0.08

C2_Incompressible

0.06

C3_Compressible

0.04
0.02
0
-0.02
-0.04
-0.06
-0.08
0

0.5

1.5

2.5

3.5

4.5

Time (s)
Figure 7.39: Influence of the dam-reservoir interaction and fluid compressibility in the
concrete cracking non-linear response. Dams crest displacement X-direction.

141

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

Velocity X direction (m/s)

1.5

C1_Empty reservoir
C2_Incompressible
C3_Compressible

1
0.5
0
-0.5
-1
-1.5
0

0.5

1.5

2.5

3.5

4.5

Time (s)
Figure 7.40: Influence of the dam-reservoir interaction and fluid compressibility in the
concrete cracking non-linear response. Dams crest velocity X-direction.

Acceleration X direction (m/s2)

40

C1_Empty reservoir
C2_Incompressible
C3_Compressible

30
20
10
0
-10
-20
-30
-40
0

0.5

1.5

2.5

3.5

4.5

Time (s)
Figure 7.41: Influence of the dam-reservoir interaction and fluid compressibility in the
concrete cracking non-linear response. Dams crest acceleration X-direction.

The consideration of the reservoir interaction in the seismic analysis is demonstrated to be important.
Fig. 7.39 shows that the maximum displacement of the full reservoir analysis (Cases C2 and C3) is
twice the maximum displacement of the empty reservoir analysis (case C1). Additionally, the vibration
period of the full reservoir cases is longer than the oscillation period of the empty reservoir case.

142

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

On the other hand, for this particular case compressibility does not play a fundamental role. Even
though higher displacement amplitudes are indeed obtained for the compressible fluid case, the
difference with the amplitudes of the incompressible fluid case is very small. Fig. 7.39 also shows that
the oscillation period of the compressible fluid case becomes progressively longer than the oscillation
period of the incompressible fluid case.
Anyhow, the time history acceleration plot shown in Fig. 7.41 demonstrates that the consideration of
the reservoir interaction and the fluid compressibility in the seismic analysis of dams produces very
different non-linear dynamic responses. This is confirmed by the cracking pattern at different time
steps shown in Fig.7.42 for each of the three case studies.

Figure 7.42: Influence of the dam-reservoir interaction and fluid compressibility in the
non-linear response. Dams concrete cracking pattern at different time steps.

Fig. 7.42 shows that when the reservoir interaction is not taken into account, the dam barely cracks;
there is only a very small cracking zone located in the dam foot. On the other hand, the first crack
appears at time step 1.42 s for the full reservoir with incompressible fluid case study. This is the case
that suffers the higher levels of cracking in the zone of the vertical dam-reservoir interface. Finally,
the higher levels of cracking in the inclined back face of the dam correspond to the compressible fluid
case study.
143

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

7. Implementation of the HFTD Method in


DIANA for the Seismic Analysis of Foundation-Dam-Reservoir Interaction Problems

144

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

8.

Conclusions and Recommendations

8. CONCLUSIONS AND RECOMMENDATIONS

8.1. Conclusions
Through the different parts of this thesis it has been demonstrated that the HFTD method is founded
on a strong theoretical basis which fundamentals is the frequency domain analysis, particularly, the
Fourier analysis.
The results of the SDOF numerical examples performed in Chapter 5 validate the high accuracy of the
HFTD method for linear and even for highly non-linear systems, including simultaneous hysteretic
stiffness and damping non-linearity, subject to several types of loading.
For linear elastic SDOF systems, the HFTD method results are almost exact for all types of loading,
with exception of sudden loading functions which starts at an initial value different than zero (f.e.:
cosine loading function). In these latter cases the results obtained are very accurate but not exact.
For non-linear SDOF systems, if the size of the time step is appropriate, the HFTD method also shows
a great precision, similar to the accuracy of the Newmark method. In order to achieve very accurate
results, both methods require that the time step size is in the order of the natural period of vibration
of the system divided over one hundred.
However, in addition to the size of the time step, the research work has shown that the accuracy of
the HFTD method depends on two numerical implementation issues: 1) Time segmentation approach;
and 2) Extended Fourier Transform period.
The division of the response time span of interest in a determined number of segments is a procedure
implemented in the HFTD method to achieve convergence in non-linear problems. This
implementation issue, denominated time segmentation approach, has proved to be very useful
especially when the system is highly non-linear, the time span of interest is long, or the system
includes frequency dependent properties. Besides, the division of the time span of interest in a higher
number of time segments improves the accuracy of the results pertaining to the initial time steps.
The extension of the Fourier Transform period by adding a band of zeros, called quiet zone, to the
loading signal is a fundamental implementation issue of the HFTD method, which makes possible to
satisfy the initial conditions and determine the transient response of the system, instead of the
steady-state response. The properly definition of the HFTD method must perform all the Fourier
Transforms using this extended period which resembles the infinite period of the analytical Fourier
Integral.
The HFTD method for MDOF systems was implemented in DIANA in combination with the Mode
Superposition method. This implementation option, explained in detail in Chapter 4, is theoretically
consistent due to the fact that the HFTD method solves a pseudo-linear equation in the frequency
domain.
The programming of the HFTD method for SDOF systems in MATLAB was essential for the
understanding, implementation and validation of the HFTD implementation for general applications in
DIANA.

145

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

8.

Conclusions and Recommendations

The application of the HFTD method is mainly interesting in MDOF systems with frequency dependent
properties, like the dam-reservoir interaction system subject to a seismic ground acceleration loading.
In Chapter 7 a case study of this system was tested in DIANA using the HFTD and Newmark methods.
For the linear elastic analysis of the dam-reservoir interaction system, the results obtained with the
HFTD and the Newmark methods show a very good agreement, in terms of accuracy and
performance. This is not the case for the concrete cracking non-linear behavior of the dam, which
HFTD and Newmark results show considerable differences.
The main reason for the differences in the results when dealing with the concrete cracking nonlinearity of this particular case study is the great influence that the Mode Superposition method has in
the description of the non-linear behavior of the structure determined with the HFTD method. Due to
the combination these two methods, the HFTD non-linear response of the structure is calculated
based on the superposition of the linear elastic modes of vibration, and therefore, the distribution of
higher stresses and cracking zones is ruled by their associated mode shapes.
A totally different situation is defined by the Newmark method, which solution is not based on the
Mode Superposition method, but on the solution of the complete system. Logically, the consequences
of this inherent difference between both methods cannot be evaluated with the SDOF systems studied
in Chapter 5, which total response is exactly the same as the modal response.
Therefore, one important conclusion of the research work is that the best performance of the HFTD
method, in combination with the Mode Superposition method, is obtained when it is applied to MDOF
systems with frequency dependent properties and linear or mild non-linear behavior.
Mild non-linear behavior is understood as a type of slight non-linearity which is not concentrated in
particular zones but spread in the whole structure. It is assumed that mild non-linearity does not
modify drastically the structural configuration of the linear elastic system, and consequently, the
superposition of the modal responses system can still provide a good estimation of the non-linear
response.
In the particular application of HFTD in dam-reservoir interaction systems, the frequency dependent
properties are composed by the fluid compressibility and the radiation boundary conditions of the
reservoir, like for example, the bottom absorption and the infinite extent.
From the thesis it can be concluded that these frequency dependent properties actually have an
impact in the response of the dam-reservoir interaction system. However, it was not possible to
outline any definitive conclusion about the relative importance of their inclusion in the systems
response.
Nevertheless, it is clear that the radiation boundary conditions, especially the bottom absorption, act
like additional damping in the system, and therefore, they reduce the magnitude of the systems
response.
The effect of fluid compressibility, on the other hand, is less predictable. For linear elastic frequency
domain analysis, the assumption of a compressible fluid clearly increases the amplitudes of the
response. However, for this particular case study, the time domain response obtained with the HFTD
method shows that, even though fluid compressibility indeed produces higher peaks, this increment is
small and constitutes a minor change in the time-history response of the system.
One important advantage of the HFTD method is its numerical stability. Accurate solutions are
obtained without convergence problems mainly because the solution of the pseudo-linear equation in
146

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

8.

Conclusions and Recommendations

the frequency domain does not require the calculation of a tangent stiffness matrix for each iteration
or each time step, as it is the case of the Newton-Raphson method. On the contrary, the HFTD
method uses the same linear elastic matrices through all the process, changing only the values of the
excitation frequencies, the excitation forces and the pseudo force vector. This allows the HFTD
method to solve large 3D models achieving convergence in an efficient way.
As a final conclusion, the HFTD method shows promising perspectives for its application in the
academic and professional practice, especially for the seismic analysis of fluid-structure and soilstructure interaction systems, or any other area for which the research of dynamic non-linear behavior
of frequency dependent systems becomes every time more important.

8.2. Recommendations
It is recommended that the current implementation of the HFTD method in DIANA be used in the
investigation of different types of frequency dependent systems.
Three-dimensional models of dam-reservoir interaction systems could be elaborated not only to keep
evaluating the performance of the HFTD method, but to study more deeply the importance of fluid
compressibility and radiation boundary conditions in the time response of this kind of systems.
It would be important to test the HFTD method using different types of non-linear behavior (concrete
crushing, steel yielding, base sliding), and evaluate the results obtained in the context of the mildnon-linearity concept and the influence that the Mode Superposition method has on the HFTD
method.
The effect that different types of structural geometries have in the HFTD non-linear responses
constitutes an interest research field which has not been studied yet. For some types of structural
geometries, like for example slender structures, the principal mode shapes describe more precisely the
actual deformation and stress distribution of the system, even after the incursion in the non-linear
behavior stage.
Finally, for future research it is recommended to study the feasibility, in terms of computational
efficiency, of implementing the HFTD method without including the Mode Superposition method. This
would allow overcoming the restriction to mild non-linear behavior, but assuming the cost of
increasing drastically the size of the system of equations to be solved.

147

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

8.

Conclusions and Recommendations

148

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

References

REFERENCES

1.

2.
3.
4.

5.

6.
7.

8.

9.
10.

11.

12.
13.
14.
15.

16.
17.
18.
19.

Aprile, A., Benedetti, A., and Trombetti, T. (1994). On non-linear dynamic analysis in the
frequency domain: algorithms and applications . Earthquake Engineering and Structural
Dynamics, Vol.23, pp. 363-388.
Clough, R.W., and Penzien, J. (2003). Structural dynamics. Computers and Structures,
California, Berkeley.
Craig, R.R., and Kurdila, A.J. (2006). Fundamentals of structural dynamics. John Wiley and
Sons, New Jersey.
Chavez, J.W., and Fenves, G.L. (1993). Earthquake analysis and response of concrete gravity
dams including base sliding. Report UBC/EERC-93/07, Earthquake Eng. Research Center, Univ.
of California, Berkeley.
Chavez, J.W., and Fenves, G.L. (1994). EAG-SLIDE: A computer program for the earthquake
analysis of concrete gravity damsincluding base sliding. Report UBC/SEMM-94/02, Structural
Engineering Mechanics and Materials, Department of Civil Engineering, Univ. of California,
Berkeley.
Chopra, A.K. (2007). Dynamics of structures: theory and applications to earthquake
engineering. Third edition. Pearson Education Inc., New Jersey.
Darbre, G.R., and Wolf, J.P. (1988). Criterion of stability and implementation issues of hybrid
frequency-time domain procedure for nonlinear dynamic analysis. Earthquake Engineering and
Structural Dynamics, Vol.16, pp. 569-581.
Darbre, G.R. (1990). Seismic analysis of a non-linearly base-isolated soil-structure interacting
reactor building by way of the hybrid frequency-time domain procedure. Earthq. Eng. and
Struct. Dyn., Vol.19, pp. 725-738.
Darbre, G.R. (1996). Nonlinear dam-reservoir interaction analysis. Proc. Eleven World Conf. on
Earthquake Engineering, Acapulco, Paper No. 760.
Fenves, G.L., and Chavez, J.W. (1990). Hybrid frequency-time domain analysis of nonlinear
fluid-structure systems. Proc. Fourth U.S. Nat. Conf. on Earthquake Engineering, California, pp.
97-104.
Fenves, G.L., and Chopra, A.K. (1983). Effects of reservoir bottom absorption on earthquake
response of concrete gravity dams. Earthquake Engineering and Structural Dynamics, Vol.11,
pp. 809-829.
Fenves, G.L., and Chopra, A.K. (1984). Earthquake analysis and response of concrete gravity
dams. Report UBC/EERC-84/10, Earthquake Eng. Research Center, Univ. of California, Berkeley.
Humar, J.L. (2002). Dynamics of structures. Second edition. Balkema Publishers, Lisse.
Mansur, W.J. et al. (2000). Time-segmented frequency-domain analysis for non-linear multidegree-of-freedom structural systems. Journal of Sound and Vibration, Vol. 237(3), 457-475.
Metrikine, A.V., and Vrouwenvelder, A.C. (2006). Dynamics of structures Part 2: Wave
dynamics. Lecture Notes of Course CIE4140, Section of Structural Mechanics, Faculty of Civil
Engineering and Geosciences, Delft University of Technology, Delft.
Paz, M., and Leigh, W. (2004). Structural dynamics: theory and computation. Fifth edition.
Springer Science+Bussiness media, New York.
Rizos, D.C., and Karabalis, D.L. (2000). Fluid-soil-structure interaction. International series on
advances in earthquake engineering. WIT Press, UK.
The Mathworks, Inc. (2012). MATLAB and Statistics Toolbox Release 2012b, Natick,
Massachusetts.
TNO DIANA BV (2011). DIANA Users Manual Release 9.4.4, Delft.

149

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

20.
21.
22.
23.

References

Veletsos, A.S., and Ventura, C.E. (1985). Dynamic analysis of structures by the DFT method. J.
Struct. Eng. ASCE, Vol.111, 2625-2642.
Wilson, E.L. (2002). Three-dimensional static and dynamic analysis of structures. Third Edition.
Computers and Structures Inc., California.
Zienkiewicz, O.C., and Bettes, P. (1978). Fluid-structure dynamic interaction and wave forces.
An introduction to numerical treatment. Int J Numer Meth Eng, Vol. 13, 1-16.
Zienkiewicz, O.C., Taylor, R.L., and Zhu, J.Z. (2005). The finite element method Its basis and
fundamentals. Sixth Edition. Elsevier Butterworth-Heinemann, London.

150

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix A: Flowchart of the HFTD Method Implementation in Diana

APPENDIX A: FLOWCHART OF THE HFTD METHOD IMPLEMENTATION IN DIANA

A.1. Main HFTD flowchart

START

Set up of fluid element matrices


,

Set up of structural element matrices


,

Eigenvalue analysis of the pseudo-linear system


[

NSDOF = Number of structural degrees of freedom


NMODES = Number of modes included in the analysis
% Eigenvectors matrix (NSDOF x NMODES)
% Generalized structural mass
% Generalized structural damping
% Generalized structural stiffness
% Participation factor of mode p
% Translational X-direction participation factor of mode p
(

% Effective mass of mode p in the translational X-direction

HFRESP

FINISH

151

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix A: Flowchart of the HFTD Method Implementation in Diana

A.2. HRESP function flowchart

START

Ground acceleration signal information


NFACT = Number of factors of the ground acceleration signal
LODTIM = Values of time
DT = (LODTIM(NFACT) - LODTIM(1)) / (NFACT-1)
% Time step size
TACSIG = NFACT * DT
% Ground acceleration signal duration

Fourier period
NLODTI = Total number of factors for the Fourier period
f1 = Free vibration fundamental frequency of the structure
= Damping ratio of the structure
TFOURI = max {TACSIG + 0.75/(f1 * ); 50 / f1}
IF (MOD (TFOURI, DT) .NE. 0 ) THEN
NLODTI = (TFOURI / DT) +1
ELSE
NLODTI = TFOURI / DT
END IF
TFOURI = NLODTI * DT

% Fourier period

Extension of ground acceleration signal


DO 100, I = NFACT+1, NLODTI
LODTIM (I) = LODTIM (1) + (I-1) * DT
100
CONTINUE

Time span of interest


NFRESP = Total number of factors for the response duration
TRESP = TACSIG + (0.5 / f1)
IF (MOD (TRESP, DT) .NE. 0 ) THEN
NFRESP = (TRESP / DT) +1
ELSE
NFRESP = TRESP / DT
END IF
TRESP = NFRESP* DT

% Time span of interest

Number of time segments


NTIMSE = Number of time segments
NSAMPL = Number of samples per time segment
IF (MOD (NFRESP, NSAMPL) .NE. 0 ) THEN
NTIMSE = (NFRESP / NSAMPL) +1
ELSE
NTIMSE = NFRESP / NSAMPL
END IF

152

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix A: Flowchart of the HFTD Method Implementation in Diana

Number of excitation frequencies


MEXCIT = Maximum number of excitation frequencies
IF (MOD (NLODTI, 2) .EQ. 0 ) THEN
MEXCIT = 1 + NLODTI / 2
ELSE
MEXCIT = 1 + ( NLODTI - 1) / 2
END IF

Set up excitation frequency values


EXCITA = Sampling frequency values
FSAMPL = (NLODTI - 1) / (LODTIM (NLODTI) - LODTIM (1))
DELTAF = FSAMPL / NLODTI

% Maximum sampling frequency


% Frequency increment

DO 100, IEXCIT = 1, MEXCIT


EXCITA (IEXCIT) = (IEXCIT-1) * DELTAF
100
CONTINUE

HIGHPASS / LOWPASS identification

Interaction matrix

IEXCIT = MAX (2, HIPASS), LOPASS


Frequency dependent structural matrix

Effective fluid matrix


[

FREQUENCY LOOP

= Effective fluid matrix


[
[

]
]

% Real part
% Imaginary part

Effective added mass and damping matrices

]
[

% Real part
% Imaginary part

Effective seismic load matrix

153

FREQUENCY LOOP

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix A: Flowchart of the HFTD Method Implementation in Diana

Generalized frequency dependent matrices

YES

% Generalized effective added mass


% Generalized frequency dependent structural damping
% Generalized effective added damping
% Generalized effective seismic load

NEXT EXCITATION
FREQUENCY?

NO
Initialize response and pseudo force vectors
DISPLA = 0
VELOCI = 0
ACCELE = 0
PSEUFO = 0
GENPSF = 0

% Relative displacement of the structure in physical coordinates (NSDOF x NFRESP)


[
% Relative velocity of the structure in physical coordinates (NSDOF x NFRESP)
[
% Relative acceleration of the structure in physical coordinates (NSDOF x NFRESP)
[
% Pseudo force vector (NSDOF x NLODTI)
[
]
% Generalized pseudo force vector (NMODES x NLODTI)
[

]
]
]

DOTISE

FINISH

154

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix A: Flowchart of the HFTD Method Implementation in Diana

A.3. DOTISE function flowchart

START
ITIMSE = 1, NTIMSE

Time segment information


FPTISE = (ITIMSE - 1) * NSAMPL +1
LPTISE = ITIMSE * NSAMPL

% First point of the current time segment


% Last point of the current time segment

Get load factors for Fourier transform of ground acceleration


NLODIR = Number of loading directions
LODFAC = Values of ground acceleration signal (NLODIR x NLODTI)
SEGFAC = Values of ground acceleration for the current segment (NLODIR x NLODTI)
ACDCYF = Values of the decay function for ground acceleration (NLODIR x NDCAYF)
NDCAYF = Number of time steps for decay function extension

TIME SEGMENT LOOP

TDCAYF = NDCAYF * DT

[
[
% NDCAYF NSAMPL

]
]
]

% Decay function duration

DO 100, ISEGFC = 1, NLODTI


IF ( ISEGFC .LE. LPTISE ) THEN
SEGFAC (ISEGFC) = LODFAC (ISEGFC)
ELSE IF ( ISEGFC .LE. (LPTISE + NDCAYF) ) THEN
SEGFAC (ISEGFC) = ACDCYF (ISEGFC - LPTISE, SEGFAC (LPTISE), NDCAYF, DT)
ELSE
SEGFAC (ISEGFC) = 0
END IF
100
CONTINUE

Fourier transformation of ground acceleration

FOULOD = Values of ground acceleration Fourier transform (NLODIR x MEXCIT)

TIMSEG

Output calculation
Dynamic fluid pressures
Strains and stresses of linear elements

YES

NEXT TIME
SEGMENT?

NO
FINISH

155

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix A: Flowchart of the HFTD Method Implementation in Diana

A.4. TIMSEG function flowchart

START

Initialize variables for convergence check


GPFOLD = 0

% Generalized pseudo force values of the previous iteration (NMODES x NFRESP)

CONVTI = FPTISE - 1

% Last converged time point number

IMODE = 1, NMODES
Add decay function to the generalized pseudo force vector

MODE LOOP

PFDCYF =

Values of the decay function for generalized pseudo force (NMODES x NDCAYF)
[
]

DO 100, ISEGFC = LPTISE + 1, LPTISE + NDCAYF


GENPSF (IMODE, ISEGFC) = PFDCYF (ISEGFC - LPTISE, GENPSF (LPTISE), NDCAYF, DT)
100
CONTINUE

Fourier transformation of generalized pseudo force

CONVERGENCE LOOP

FOUGPF =

Values of generalized pseudo force Fourier transform (NMODES x MEXCIT)

[
]

YES

IEXCIT = MAX (2, HIPASS), LOPASS

NEXT MODE?

NO

FREQUENCY LOOP

Solution of pseudo linear system and determination of structural response in the frequency
domain
[

GDISFR = Generalized displacement in freq. domain (NMODES x MEXCIT)


GVELFR = Generalized velocity in freq. domain (NMODES x MEXCIT)
GACCFR = Generalized acceleration in freq. domain (NMODES x MEXCIT)

YES

NEXT EXCITATION
FREQUENCY?

NO

156

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix A: Flowchart of the HFTD Method Implementation in Diana

IMODE = 1, NMODES

MODE LOOP

Inverse Fourier transformation of generalized response

GDISTI = Generalized disp. in time domain (NMODES x NLODTI)


GVELTI = Generalized velocity in time domain (NMODES x NLODTI)
GACCTI = Generalized accel. in time domain (NMODES x NLODTI)

YES

[
[
[

]
]
]

NEXT MODE?

NO

ITIME = CONVTI + 1, LPTISE

Expand from generalized to physical response

NOT CONVERGED TIME STEPS LOOP

CONVERGENCE LOOP

DISPLA (ITIME)
VELOCI (ITIME)
ACCELE (ITIME)

Nonlinear internal restoring force (State space determination)


FNONLI (ITIME)

Linear internal restoring force


FLINEA (ITIME)

Pseudo force in physical and generalized coordinates


PSEUFO (ITIME)
GENPSF (ITIME)

Calculate error
GPSFER =

Error of the generalized pseudo force vector in relation with


the previous iteration

IF ( GENPSF(ITIME) .EQ. 0 ) THEN


GPSFER = 0
ELSE
GPSFER = |GENPSF (ITIME) GPFOLD (ITIME)| / |GENPSF (ITIME)|

Update the last iteration generalized pseudo force


GPFOLD (ITIME) = GENPSF (ITIME)

157

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix A: Flowchart of the HFTD Method Implementation in Diana

NOT CONVERGED TIME STEPS LOOP

CONVERGENCE LOOP

GPSFER GPSTOL
&
CONVTI = ITIME - 1

NO

YES
Update the converged time step counter
CONVTI = ITIME

YES

NEXT TIME
STEP?

NO
NO

CONVTI = LPTISE

YES

FINISH

158

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

APPENDIX B: MATLAB ROUTINES

B.1. HFTD method code

B.1.1. Main HFTD script

Section 1: File name ........................................................................................................................................... 160


Section 2: Exercise data definition .................................................................................................................... 161
Section 3: Construction of the linear elastic matrices of the system ........................................................... 164
Section 4: Free vibration Eigen-analysis .......................................................................................................... 165
Section 5: Fundamental mode of vibration damping ratio ............................................................................ 166
Section 6: Ground acceleration signal information ......................................................................................... 167
Section 7: Time step size and Fourier Period .................................................................................................. 168
Section 8: Extension of ground acceleration signal ........................................................................................ 169
Section 9: Time span of interest ....................................................................................................................... 170
Section 10: Number of time segments ............................................................................................................. 171
Section 11: Decay function length .................................................................................................................... 172
Section 12: Set up excitation frequencies ....................................................................................................... 173
Section 13: Highpass/Lowpass frequency identification ................................................................................ 174
Section 14: Hysteretic and frequency dependent damping for each excitation frequency ...................... 175
Section 15: Initialization of variables................................................................................................................ 175
Section 16: HFTD method execution for each time segment ....................................................................... 176
Section 17: Output results post processing ..................................................................................................... 183

159

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
This script execute the HFTD method for linear and nonlinear simplfied
%
dynamic models of one or more degrees of freedom (DOF) subjected to
%
base (ground) acceleration loading.
%
%
After defining the system properties, the load signal and the HFTD
%
method parameters, the following results are obtained: time history
%
displacement, velocity, acceleration and nonlinear internal forces of
%
the system. Besides, the frequency response of the pseudo linear system
%
is also obtained.
%
%
The ground acceleration signal should be saved with the name "signal"
%
in the work folder as a matlab variable file called "signal.mat". The
%
values of the ground acceleration stored in the variable signal should
%
be normalized with respect to the gravity (g).
%
%
A Frequency dependent damping relation for the dashpots coefficients
%
can be specified in the function "f2_freq_dep_damp1".
%
%
The desired type of decaying function should be defined in the
%
function file "f5_decay_func".
%
%
The type of stiffness and damping non-linearity to be used in the
%
problem should be selected in Section 16.6.6 of this script.
%
The selected function file is used to determine FNONLI.
%
The nonlinear parameters of each type of non-linear function should be
%
defined inside the corresponding function file.
%
%
The output results are saved as excel tables, figures and matlab data
%
files. The file path where these output results files are saved should
%
be indicated Secttion 17 of this script and in the function files
%
"f15_excel_tables_v2" and "f16_plot_resp_v2".
%
%

Section 1: File name


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
All the output results of the program will be saved in the folder
%
"Output results" located in the work folder. The name specified by
%
the user is used to save the output files.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
clear;

160

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

disp('==========================================================================')
disp('
Problem name
')
disp('==========================================================================')
name=input('\nEnter the name of the file (this name is used to save the output data)... ','s');
disp('
')
pause(1);

Section 2: Exercise data definition


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Input the mass, damping and stiffness matrices of the simplified MDOF
%
system. The gravity constant and the tolerance for non-linear analysis
%
are also input by the user.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Definition of the mass matrix
')
disp('==========================================================================')
pause(1);
lock=0;
while lock==0
MCOF=input('\nEnter the value of the translational masses for each degree \nof freedom m (in
a row vector enter one mass per DOF).... ');
sizem=size(MCOF);
if sizem(1)~=1
disp(' ')
disp('Error message: The masses should be arranged in a row vector')
pause(2);
else
lock=1;
end
end
NDOF=sizem(2);
pause(1);
disp(' ')
disp('The number of degrees of freedom (NDOF) of the system is defined to be ')
NDOF
pause(1)
disp(' ')
disp('The masses of each degree of freedom are defined to be ')
MCOF
pause(1);
disp(' ')
disp('==========================================================================')
disp('
Definition of the elastic stiffness matrix
')
disp('==========================================================================')
pause(1);
lock=0;
while lock==0

161

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

KELAST=input('\nEnter the linear elastic stiffness coefficients k of the springs \n(in a row
vector enter NDOF+1 stiffness coefficients) .... ');
sizek=size(KELAST);
if sizek(1)~=1
disp(' ')
disp('Error message: The spring coefficients should be arranged in a row vector')
pause(2);
elseif sizek(2)~=NDOF+1
disp(' ')
disp('Error message: The row vectors defining the coefficientes of K should have NDOF+1
components ')
pause(2);
else
lock=1;
end
end
disp(' ')
disp('The linear elastic stiffness coefficients are defined to be')
KELAST
disp(' ')
pause(1);
disp('==========================================================================')
disp('
Definition of the damping matrices
')
disp('==========================================================================')
pause(1);
lock=0;
while lock==0
CELAST=input('\nEnter the constant (not frequency dependent) linear elastic damping
coefficients c \nof the dashpots (in a row vector enter NDOF+1 damping coefficients)... ');
sizec=size(CELAST);
if sizec(1)~=1
disp(' ')
disp('Error message: The dashpots coefficients should be arranged in a row vector')
pause(2);
elseif sizec(2)~=NDOF+1
disp(' ')
disp('Error message: The row vectors defining the constant dashpot coefficients should
have NDOF+1 components ')
pause(2);
else
lock=1;
end
end
disp(' ')
disp('The constant (not frequency dependent) linear elastic damping coefficients are defined to
be ')
CELAST
pause(1);
lock=0;
while lock==0

162

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

RAYLEI=input('\nEnter the coeficients for the definition of Rayleigh damping matrix aM+bK
\nin a row vector [a b]. If there is no Rayleigh damping enter [0 0]... ');
sizer=size(RAYLEI);
if sizer(1)~=1
disp(' ')
disp('Error message: The Rayleigh coefficients should be arranged in a row vector')
pause(2);
elseif sizer(2)~=2
disp(' ')
disp('Error message: The row vectors defining the Rayleigh coefficientes a and b should
have 2 components ')
pause(2);
else
lock=1;
end
end
a=RAYLEI(1);
b=RAYLEI(2);
disp(' ')
disp('The Rayleigh damping coefficients are defined to be ')
a
b
pause(1);
lock=0;
while lock==0
HYSTER=input('\nEnter the hysteretic damping ratio. If there is no hysteretic damping enter
0... ');
sizeh=size(HYSTER);
if sizeh(1)==1 && sizeh(2)==1
lock=1;
else
disp(' ')
disp('Error message: The input value must be scalar')
end
end
disp(' ')
disp('The hysteretic damping ratio is defined to be ')
HYSTER
disp(' ')
pause(1);
disp('==========================================================================')
disp('
Gravity constant
')
disp('==========================================================================')
pause(1);
lock=0;
while lock==0
g=input('\nEnter the value of the gravity constant g .... ');
sizeg=size(g);
if sizeg(1)~= 1 || sizeg(2)~=1
disp(' ')

163

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

disp('Error message: g must be a scalar value')


pause(2);
else
lock=1;
end
end
disp(' ')
disp('The gravity constant is defined to be ')
g
disp('
')
pause(1);
disp('==========================================================================')
disp('
Tolerance for convergence of non-linear analysis
')
disp('==========================================================================')
pause(1);
lock=0;
while lock==0
tol=input('\nEnter the convergence tolerance tol for non-linear analysis.... ');
sizet=size(tol);
if sizet(1)~= 1 || sizet(2)~=1
disp(' ')
disp('Error message: tol must be a scalar value')
pause(2);
elseif tol>0.01 || tol<=0
disp(' ')
disp('Error message: tol must be positive and smaller than 10^-2')
pause(2);
else
lock=1;
end
end
disp(' ')
disp('The tolerance for convergence check is defined to be ')
tol
pause(2);
disp(' ')

Section 3: Construction of the linear elastic matrices of the system


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Based on the coefficients values defined in the previous section for
%
each mass, spring and dashpot, the linear elastic matrices of the
%
system are constructed using the function file "f1_linelast_mat".
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Linear elastic matrices
')
disp('==========================================================================')
%Mass matrix

164

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

M=zeros(NDOF);
for IDOF=1:NDOF
M(IDOF,IDOF)=MCOF(IDOF);
end
disp(' ')
disp('The mass matrix M is')
M
pause(1);
%Stiffness matrix
K=f1_linelast_mat(NDOF,KELAST);
disp(' ')
disp('The linear elastic stiffness matrix K is')
K
pause(1);
%Damping matrix
Cnp=f1_linelast_mat(NDOF,CELAST);
disp(' ')
disp('The constant (not frequency dependent) linear elastic non proportional (dashpots) damping
matrix Cnp is')
Cnp
pause(1)
Cray=a*M+b*K;
disp(' ')
disp('The constant (not frequency dependent) linear elastic Rayleigh damping matrix Cray is')
Cray
pause(1)
C=Cnp+Cray;
disp(' ')
disp('The total constant (not frequency dependent) linear elastic damping matrix C is')
C
pause(1)
Ch=HYSTER*K;
disp(' ')
disp('The hysteretic linear elastic damping matrix Ch is')
Ch
pause(2);
disp(' ')

Section 4: Free vibration Eigen-analysis


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Eigen-analysis of the simplified system. Mode shape vectors and
%
eigen-frequencies of the system are calculated.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Free vibration natural frequencies and mode shape vectors
disp('==========================================================================')
[Phi,Natfreq]=eig(K,M);

')

165

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

Mgen=transpose(Phi)*M*Phi;
w1=sqrt(Natfreq(1,1));
T1=2*pi/w1;
pause(1);
disp('
')
disp('The eigenvectors matrix is ')
disp('
')
Phi
pause(1);
disp('
')
disp('The eigenvalues (squared natural circular frequencies) matrix is')
disp('
')
Natfreq
pause(1);
disp('
')
disp('The fundamental period of the system is defined to be')
disp('
')
T1
pause(1);
disp('
')
disp('The fundamental circular frequency of the system is defined to be')
disp('
')
w1
pause(1);

Section 5: Fundamental mode of vibration damping ratio


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Damping ratio of the fundamental mode of vibration taking into account
%
frecuency dependent and hysteretic damping. The frequency dependent
%
damping should be defined by the user in the function file
%
"f2_freq_dep_damp1".
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Damping ratio of the fundamental mode of vibration considering
')
disp('
hysteretic and frequency dependent damping
')
disp('==========================================================================')
[CFREDE1,EM]=f2_freq_dep_damp1(NDOF,w1);% Function defining the frequency dependent damping
coefficients
if EM==1
return
end
Cfd1=f1_linelast_mat(NDOF,CFREDE1);
Ch1=(1/w1)*Ch;
pause(1);
disp('
')
disp('The hysteretic damping matrix (Ch1) and the frequency dependent damping ')
disp('matrix (Cfd1) corresponding to the fundamental mode of vibration are defined to be ')

166

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

Ch1
Cfd1
pause(1);
C1=C+Ch1+Cfd1;
disp(' ')
disp('The total damping matrix (constant + hysteretic + frequency dependent) ')
disp('for the fundamental mode of vibration is determined to be ')
C1
pause(1);
C1gen=transpose(Phi(:,1))*C1*Phi(:,1);
xi1=(1/(2*w1))*(C1gen/Mgen(1,1));
disp('
')
disp('The damping ratio of the fundamental mode of vibration of the system is defined to be')
disp('
')
xi1
disp('
')
pause(2);

Section 6: Ground acceleration signal information


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The data of the ground acceleration signal must be stored as a
%
matlab variable called "signal". This variable should be
%
defined as a matrix with two columns and as many rows as time
%
points available in the signal data. The first column contains the
%
time steps values, which should start in time = 0 seconds, and
%
should be equally spaced (constant time step size). The second
%
colum contains the ground acceleration values normalized with the
%
gravity acceleration g.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Ground acceleration signal information
')
disp('==========================================================================')
disp('
')
disp('Loading the ground acceleration variable "signal" from file signal.mat saved in the work
folder...')
disp('
')
pause(1);
load signal;
lock=1;
while lock~=0
sizesig=size(signal);
[SIGTIM,NFACT,SIGNDT,lock,k]=f3_ground_accel_signal_check(signal,sizesig);% Function that
checks the correct definition of the signal variable
if lock==1
disp('Error message: The first time step must be zero')
return
elseif lock==2

167

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

disp('Error message: Signal data variable must have two columns')


return
elseif lock==3
disp('Error message: Time step size of signal data is not constant')
disp('
')
pause(2);
fprintf('The problem was found between the time points %d and %d\n',k,k+1)
return
end
end
SIGFAC=g*signal(:,2);
TACSIG=NFACT*SIGNDT;
pause(2);
disp('
')
disp('The number of time points contained in the ground acceleration signal is ')
disp('
')
NFACT
pause(1);
disp('
')
disp('The time step size of the ground acceleration signal is ')
disp('
')
SIGNDT
pause(1);
disp('
')
disp('The duration of the ground acceleration signal is ')
disp('
')
TACSIG
pause(1);
disp('
')
disp('The maximum ground acceleration magnitude contained in the signal is ')
disp('
')
disp(max(abs(SIGFAC)))
disp('
')
pause(2);

Section 7: Time step size and Fourier Period


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The time step size (DT) is calculated base on the signal
%
discretization.
%
The Fourier period length (TFOURI)is calculated based on the
%
fundamental mode of vibration period (T1) and damping ratio (xi1).
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('==========================================================================')
disp('
Period for the Fourier transform
')
disp('==========================================================================')
lock=0;
while lock==0
pause(2);

168

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

opt=input('\nSelect the option for the definition of the HFTD time step size (DT). Enter "0"
for DT to be \nequal to the signal step size (SIGNDT) or enter "1" for DT to be a multiple of
SIGNDT .... ');
if opt==0
TIMES=1;
lock=1;
elseif opt==1
pause(2);
TIMES=input('\nEnter the number of times that SIGNDT should be multiplied to obtain
DT.... ');
if mod(TIMES,1)~=0 || TIMES<=0
disp('
')
disp('Error message: The input value should be a positive integer')
else
lock=1;
end
else
pause(2);
disp('
')
disp('Error message: The input value must be 0 or 1')
end
end
DT=TIMES*SIGNDT;
prob=[TACSIG+0.75*T1/abs(xi1),50*T1,2*TACSIG];
TFOURI=max(prob);
NLODTI=ceil(TFOURI/DT);
TFOURI=DT*NLODTI;
pause(1);
disp('
')
disp('The number of time points (NLODTI) contained in the Fourier period is calculated to be ')
disp('
')
NLODTI
pause(1);
disp('
')
disp('The HFTD time step size (DT) is defined to be')
disp('
')
DT
pause(1);
disp('
')
disp('The Fourier period (TFOURI) is calculated to be')
disp('
')
TFOURI
disp('
')
pause(2)

Section 8: Extension of ground acceleration signal


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The ground acceleration data is extended adding zeros (quiet zone)

169

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%
until the Fourier period is covered.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LODTIM=zeros(NLODTI,1);
LODFAC=zeros(NLODTI,1);
for k=1:NLODTI
LODTIM(k)=(k-1)*DT;
j=1+(k-1)*TIMES;
if j<=NFACT
LODFAC(k)=SIGFAC(j);
else
LODFAC(k)=0;
end
end

Section 9: Time span of interest


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Time span for which the reponses are calculated.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('==========================================================================')
disp('
Time span of interest
')
disp('==========================================================================')
lock=0;
while lock==0
pause(2);
opt=input('\nSelect the option for the definition of the time span of interest: Enter "1" for
the time span of interest to be \ndefined by the user or enter "0" for it to be calculated by te
program .... ');
if opt==0
TRESP=TACSIG+0.5*T1;
NFRESP=ceil(TRESP/DT);
TRESP=DT*NFRESP;
lock=1;
elseif opt==1
pause(1);
TRESP=input('\nEnter the time span of interest in seconds .... ');
NFRESP=ceil(TRESP/DT);
TRESP=DT*NFRESP;
if TRESP>=TFOURI/2 || TRESP<=0
disp('
')
disp('Error message: The time span of interest should be posistive and not greater
than half the Fourier period (TFOURI)')
else
lock=1;
end
else
pause(1);
disp('
')
disp('Error message: The input value must be 0 or 1')

170

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

end
end
disp('
disp('The
be')
disp('
NFRESP
pause(1);
disp('
disp('The
disp('
TRESP
disp('
pause(2);

Appendix B: MATLAB Routines

')
number of time steps defining the time span of interest length (NFRESP) is defined to
')

')
response time span of interest (TRESP) is defined to be')
')
')

Section 10: Number of time segments


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
In order to achieve accurate results, HFTD requires that the time
%
span of interest be divided in segments containing a certain number
%
(NSAMPL) of time steps (DT). The convergence and iterative process
%
is carried out sequentially, segment by segment, one after the
%
other.
%
One extreme case (not recommended) is to have only one time segment
%
(NSAMPL*DT=TRESP). On the other hand, the other extreme case is to
%
have as many segements as time steps (NSAMPL=1). Intermediate case
%
is reccommended taking into account that increasing the number of
%
time segments provides better accuracy and stability to the HFTD
%
method.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('==========================================================================')
disp('
Division of the time span of interest in time segments
')
disp('==========================================================================')
lock=0;
while lock==0
pause(2);
NSAMPL=input('\nEnter the number of time steps (DT) per time segment .... ');
if mod(NSAMPL,1)==0 && NSAMPL>=1 && NSAMPL<=NFRESP
lock=1;
else
pause(1);
fprintf('\nError message: The number of samples per time segment (NSAMPL) should be a
positive \ninteger such that NSAMPL*DT is not greater than TRESP')
end
end
disp('
')
disp('The number of samples per time segment (NSAMPL) is defined to be')
disp('
')
NSAMPL

171

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

disp('
')
NTIMSE=ceil(NFRESP/NSAMPL);
pause(1);
disp('The number of time segments (NTIMSE) through all the interest time span is defined to be')
disp('
')
NTIMSE
disp('
')
pause(2);

Section 11: Decay function length


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
For the Fourier transformation of the ground acceleration and the
%
pseudo force vector, an decaying function is appended at the end of
%
the current time segment being analyzed.
%
The number of time steps (DT) that define the decay function length
%
is stored in the variable NDCAYF, which should be smaller or equal
%
than NSAMPL (the decay function should not extend more than the
%
time segment length).
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('
')
disp('==========================================================================')
disp('
Number of time steps defining the decaying function length
')
disp('==========================================================================')
lock=0;
while lock==0
pause(2);
NDCAYF=input('\nEnter the number of time steps (DT) for the definition of the decaying
function length .... ');
if NDCAYF>=0 && mod(NDCAYF,1)==0 && NDCAYF<=NSAMPL
lock=1;
else
disp('
')
fprintf('Error message: The number of time steps defining the decaying function length
(NDCAYF) should be \nzero or a positive integer not greater than NSAMPL\n')
end
end
TDCAYF=NDCAYF*DT;
disp('
')
disp('The number of time steps defining the length of the decaying function (NDCAYF) is defined
to be')
NDCAYF
pause(1)
disp('
')
disp('The lentgh of the decaying function (TDCAYF) is defined to be')
TDCAYF
disp('
')
pause(2);

172

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

Section 12: Set up excitation frequencies


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The number of excitation frequencies (MEXCITA), the frequency
%
increment (DELTAF) and the maximum sampling frequency (FSAMPL) are
%
calculated based on the Fourier Period (TFOURI) and its number of
%
time points (NLODTI).
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('===========================================================================================
=========================================')
fprintf('
Number of excitation frequencies, excitation frequency values, frequency increment \n
and maximum sampling frequency.
\n')
disp('===========================================================================================
=========================================')
if mod(NLODTI,2)==0
MEXCIT=1+NLODTI/2;
else
MEXCIT=1+(NLODTI-1)/2;
end
FSAMPL=(NLODTI-1)/(LODTIM(NLODTI)-LODTIM(1));
DELTAF=FSAMPL/NLODTI;
EXCITA=zeros(MEXCIT,1);
for IEXCIT=1:MEXCIT
EXCITA(IEXCIT)=(IEXCIT-1)*DELTAF;
end
pause(2);
disp('
')
disp('The maximum number of frequencies (MEXCIT) is')
disp('
')
MEXCIT
disp('
')
pause(1);
disp('The frequency increment (DELTAF) is')
disp('
')
DELTAF
disp('
')
pause(1);
disp('The maximum sampling frequency (FSAMPL) is')
disp('
')
FSAMPL
disp('
')
pause(1);
disp('The maximum excitation frequency that can be accurately included in the analysis is')
disp('
')
disp(FSAMPL/2)
disp('
')
pause(2);

173

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

Section 13: Highpass/Lowpass frequency identification


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The frequency response can be determined for a restricted range of
%
frequency instead of the frequency range from 0 to EXCITA(MEXCIT).
%
This restriction is (optionally) defined by the user after
%
specifying a minimum (Highpass) and a maximum (Lowpass) response
%
frequency.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('==========================================================================')
disp('
Highpass/Lowpass frequency identification
')
disp('==========================================================================')
lock=0;
while lock==0
pause(2);
MINFRE=input('\nEnter the minimum frequency to be included in the analysis .... ');
pause(1);
if MINFRE>=0 && MINFRE<=EXCITA(MEXCIT)
lock=1;
else
disp('
')
fprintf('Error message: The minimum frequency (MINFRE) should be zero or a positive value
\nnot greater than (MEXCIT-1)*DELTAF\n')
end
end
lock=0;
while lock==0
MAXFRE=input('\nEnter the maximum frequency to be included in the analysis .... ');
pause(1);
disp('
')
if MAXFRE>=MINFRE && MAXFRE<=EXCITA(MEXCIT)
lock=1;
else
fprintf('Error message: The maximum frequency (MAXFRE) should higher or equal than MINFRE
\nand not greater than (MEXCIT-1)*DELTAF\n')
end
end
[LOPASS,HIPASS]=f4_high_low_pass_freq(EXCITA,MEXCIT,MAXFRE,MINFRE);
disp('The HIGHPASS frequency is found to be')
disp('
')
disp(EXCITA(HIPASS))
disp('
')
pause(1);
disp('The LOWPASS frequency is found to be')
disp('
')
disp(EXCITA(LOPASS))

174

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

Section 14: Hysteretic and frequency dependent damping for each excitation frequency
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The hysteretic and frequency dependent damping matrices are
%
calculated for each excitattion frequency between HIPASS and
%
LOPASS. The frequency dependent damping should be defined by the
%
user in the function file "f2_freq_dep_damp1".
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
CFREDE=zeros(LOPASS-HIPASS+1,NDOF+1);
Cfd=zeros(NDOF,NDOF,LOPASS-HIPASS+1);
Chy=zeros(NDOF,NDOF,LOPASS-HIPASS+1);
for IEXCIT=HIPASS:LOPASS
w=2*pi*EXCITA(IEXCIT);
[CFREDE(IEXCIT-HIPASS+1,:),EM]=f2_freq_dep_damp1(NDOF,w); %Function defining the frequency
dependent damping
Cfd(:,:,IEXCIT-HIPASS+1)=f1_linelast_mat(NDOF,CFREDE(IEXCIT-HIPASS+1,:));
if w==0
Chy(:,:,IEXCIT-HIPASS+1)=0*Ch;
else
Chy(:,:,IEXCIT-HIPASS+1)=(1/w)*Ch;
end
end

Section 15: Initialization of variables


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Set up to zero the response, pseudo force and internal restoring
%
force variables
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
NRESIT=NTIMSE*NSAMPL;
DISPLA=zeros(NDOF,NRESIT);
VELOCI=zeros(NDOF,NRESIT);
ACCELE=zeros(NDOF,NRESIT);
PSEUFO=zeros(NDOF,NLODTI);
FNONLI=zeros(NDOF,NRESIT);
FLINEA=zeros(NDOF,NRESIT);
KEY=zeros(NDOF+1,NRESIT);
FNLSTI=zeros(NDOF,NRESIT);
SPRFOR=zeros(NDOF+1,NRESIT);
FNLDAM=zeros(NDOF,NRESIT);
DASHFO=zeros(NDOF+1,NRESIT);
UT=zeros(NDOF+1,NRESIT);
UC=zeros(NDOF+1,NRESIT);
UO=zeros(NDOF+1,NRESIT);
disp('
')
pause(1);

175

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

Section 16: HFTD method execution for each time segment


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The HFTD calculation process is executed segment by segment. A time
%
segment loop is defined in order to repeat the process for the
%
NTIMSE segments previously determined.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('==========================================================================')
disp('
HFTD method execution for each time segment
')
disp('==========================================================================')
for ITIMSE=1:NTIMSE
pause(1);
disp('
')
disp('Starting the HFTD calculation process: Time segment number ... ')
disp(ITIMSE)

Section 16.1: Time segment information


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The index of the first (FPTISE) and last (LPTISE) point of the
%
current time segment (ITIMSE) are identified.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
FPTISE=(ITIMSE-1)*NSAMPL+1;
LPTISE=ITIMSE*NSAMPL;

Section 16.2: Ground acceleration decay function definition for the current segment
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The function "f5_decay_func" is called in order to determine
%
the ground acceleration factors pertaining to the decay
%
function appended at the end of the current time segment.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
[GADCYF]=f5_decay_func(LODFAC,LPTISE,NDCAYF,DT,TDCAYF);

Section 16.3: Load factors for the Fourier transform of the ground acceleration
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The load factors used for the Fourier transform of the ground
%
acceleration vary in each time segment. The ground acceleration
%
load factors including the decaying function used for the
%
current time segment are stored in the variable SEGFAC.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
SEGFAC=LODFAC;
SEGFAC(LPTISE+1:LPTISE+NDCAYF)=GADCYF;
SEGFAC(LPTISE+NDCAYF+1:NLODTI)=zeros;

176

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

Section 16.4: Fast Fourier transform of the ground acceleration loading


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The FFT function is executed in the variable SEGFAC by
%
means of the function "f6_fast_fourier_transf".
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('Fourier transform of the ground acceleration signal ... ')
[FOUACR,FOUACI]=f6_fast_fourier_transf(SEGFAC,NLODTI,MEXCIT);
FOUACC=FOUACR+1i*FOUACI;

Section 16.5: Initialize variables for convergence check


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Set up to zero the response, pseudo force and internal
%
restoring force variables
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
PSFOLD=zeros(NDOF,NSAMPL); %Pseudo force values of the last iteration within the current
segment.
DISOLD=zeros(NDOF,NSAMPL); %Displacement values of the last iteration within the current
segment.
PFDCYF=zeros(NDOF,NDCAYF); %Decay function for the pseudo force vector.
FOUPFR=zeros(NDOF,MEXCIT); %Direct Fourier transform of the pseudo force vector (real part).
FOUPFI=zeros(NDOF,MEXCIT); %Direct Fourier transform of the pseudo force vector (imaginary
part).
IFOUDR=zeros(NDOF,NSAMPL); %Inverse Fourier transform of the displacement vector(real part)
for the time points inside the current time segment.
IFOUDI=zeros(NDOF,NSAMPL); %Inverse Fourier transform of the displacement vector(imaginary
part) for the time points inside the current time segment.
IFOUVR=zeros(NDOF,NSAMPL); %Inverse Fourier transform of the velocity vector(real part) for
the time points inside the current time segment.
IFOUVI=zeros(NDOF,NSAMPL); %Inverse Fourier transform of the velocity vector(imaginary part)
for the time points inside the current time segment.
IFOUAR=zeros(NDOF,NSAMPL); %Inverse Fourier transform of the acceleration vector(real part)
for the time points inside the current time segment.
IFOUAI=zeros(NDOF,NSAMPL); %Inverse Fourier transform of the acceleration vector(imaginary
part) for the time points inside the current time segment.
CONVTI=FPTISE-1;
%Counter of the last time steps already converged.

Section 16.6: Convergence loop


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The procedure below is repeated until achieving convergence in
%
all the time steps within the current segment.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
iter=0;
convlock=0;
pause(1);
disp('
')

177

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

fprintf('Starting the iteration process untill convergence is achieved for all the time steps
\ninside segment number... %d\n',ITIMSE)
while convlock==0
iter=iter+1;
disp('
')
disp('Starting iteration number... ')
disp(iter)
disp('
')
disp('Fourier transform of the pseudo force vector ... ')
for IDOF=1:NDOF

Section 16.6.1: Pseudo force decay function definition


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The function "f5_decay_func is called in order to
%
determine the pseudo force factors pertaining to the
%
decay function appended at the end of the current time
%
segment.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
[PFDCYF(IDOF,:)]=f5_decay_func(PSEUFO(IDOF,:),LPTISE,NDCAYF,DT,TDCAYF);

Section 16.6.2: Load factors for the Fourier transform of the pseudo force
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The load factors used for the Fourier transform of the
%
pseudo force vary in each iteration in accordance with
%
the results obtained in the previous iteration. The
%
pseudo force load factors including the decaying
%
function used for the current time segment and
%
iteration are stored in the variable PSEUFO.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
PSEUFO(IDOF,LPTISE+1:LPTISE+NDCAYF)=PFDCYF(IDOF,:);

Section 16.6.3: Fast Fourier transform of the pseudo force


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The FFT function is executed in the variable PSEUFO by
%
means of the function "f6_fast_fourier_transf".
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
[FOUPFR(IDOF,:),FOUPFI(IDOF,:)]=f6_fast_fourier_transf(PSEUFO(IDOF,:),NLODTI,MEXCIT);
end

178

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

Section 16.6.4: Pseudo linear equation of motion in the frequency domain


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Solution of the pseudo linear system for each excitation
%
frequency.
%
%
${-(w^2)*M + i*w*C + K} u(w) = -M*I*ug(w) + q(w)$
%
[-(w^2)*M + i*w*C + K] u(w) = -M*I*ug(w) + q(w)
%
The equation is solved using the direct method defined in the
%
function "f7_solver_pseudo_lin_direct_freq_v2".
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('
')
disp('Starting the solution of the pseudo linear system in the frequency domain ... ')
DISFRE=zeros(NDOF,NLODTI);
VELFRE=zeros(NDOF,NLODTI);
ACCFRE=zeros(NDOF,NLODTI);
DIFREA=zeros(NDOF,MEXCIT);
DIFRIM=zeros(NDOF,MEXCIT);
for IEXCIT=HIPASS:LOPASS
w=2*pi*EXCITA(IEXCIT);
[DIFREA(:,IEXCIT),DIFRIM(:,IEXCIT)]=f7_solver_pseudo_lin_direct_freq_v2(K,C,Cfd(:,:,IEXCITHIPASS+1),Chy(:,:,IEXCITHIPASS+1),M,w,NDOF,FOUACR(IEXCIT),FOUACI(IEXCIT),FOUPFR(:,IEXCIT),FOUPFI(:,IEXCIT));
DISFRE(:,IEXCIT)=DIFREA(:,IEXCIT)+1i*DIFRIM(:,IEXCIT); % Displacement in the
frequency domain
VELFRE(:,IEXCIT)=1i*w*DISFRE(:,IEXCIT);
% Velocity in the frequency domain
ACCFRE(:,IEXCIT)=-(w^2)*DISFRE(:,IEXCIT); % Acceleration in the frequency domain
end

Section 16.6.5: Inverse Fourier transform of the response


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The complex displacement (DISFRE), velocity (VELFRE) and
%
acceleration (ACCFRE) vectors in the frequency domain are
%
translated to the time domain by means of the function
%
"f8_inv_fast_fourier_transf".
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('
')
disp('Inverse Fourier transform of the displacement, velocity and acceleration vectors
... ')
for IDOF=1:NDOF
[IFOUDR(IDOF,:),IFOUDI(IDOF,:)]=f8_inv_fast_fourier_transf(DISFRE(IDOF,:),NLODTI,FPTISE,LPTISE);
% Inverse Fourier transform of the displacement vector for the time steps within the current
segment.
[IFOUVR(IDOF,:),IFOUVI(IDOF,:)]=f8_inv_fast_fourier_transf(VELFRE(IDOF,:),NLODTI,FPTISE,LPTISE);
% Inverse Fourier transform of the velocity vector for the time steps within the current

179

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

segment.
[IFOUAR(IDOF,:),IFOUAI(IDOF,:)]=f8_inv_fast_fourier_transf(ACCFRE(IDOF,:),NLODTI,FPTISE,LPTISE);
% Inverse Fourier transform of the acceleration vector for the time steps within the current
segment.
end

Section 16.6.6: Not converged time steps loop


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
In this loop, the new iteration values of the displacement,
%
velocity, acceleration and pseudo force vectors are
%
calculated and updated only for the time steps inside the
%
current segment which have not converged in the previous
%
iterations.
%
The functions defining the stiffness and damping
%
non-linearities are indicated below.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('
')
disp('Calculation of the response, internal restoring forces and pseudo force for the non
converged time steps ... ')
for ITIME=CONVTI+1:LPTISE
DISPLA(:,ITIME)=IFOUDR(:,ITIME-FPTISE+1); % Displacement vector
VELOCI(:,ITIME)=IFOUVR(:,ITIME-FPTISE+1); % Velocity vector
ACCELE(:,ITIME)=IFOUAR(:,ITIME-FPTISE+1); % Acceleration vector
if ITIME==1
CEROS1=zeros(NDOF,1);
CEROS2=zeros(NDOF+1,1);

Section 16.6.6.1: Non-linear or linear stiffness force, internal spring force and state variable
of time step 1

%[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f9_lin_
stiff_int_force(NDOF,CEROS2,CEROS1,DISPLA(:,ITIME),CEROS2,KELAST);
%[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f10_non
lin_hyster_stiff_int_force_v3(VELOCI(:,ITIME),NDOF,CEROS2,CEROS1,DISPLA(:,ITIME),CEROS2,CEROS2,CE
ROS2,CEROS2,ITIME,KELAST);
[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f11_nonl
in_hyster_stiff_int_force_cubic(VELOCI(:,ITIME),NDOF,CEROS2,CEROS1,DISPLA(:,ITIME),CEROS2,CEROS2,
CEROS2,ITIME,KELAST);

Section 16.6.6.2: Non-linear or linear damping force, internal dashpot force and state
variable of time step 1
180

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%[FNLDAM(:,ITIME),DASHFO(:,ITIME)]=f12_lin_damp_int_force_v2(NDOF,VELOCI(:,ITIME),CEROS2,CELAST);
%Nonlinear damping force and internal dashpot force of time step i+1
[FNLDAM(:,ITIME),DASHFO(:,ITIME)]=f13_nonlinmet_damp_int_force(NDOF,DISPLA(:,ITIME),VELOCI(:,ITIM
E),CELAST);
else

Section 16.6.6.3: Non-linear or linear stiffness force, internal spring force and state variable
of time step i+1

%[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f9_lin_
stiff_int_force(NDOF,KEY(:,ITIME-1),DISPLA(:,ITIME-1),DISPLA(:,ITIME),SPRFOR(:,ITIME-1),KELAST);
%[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f10_non
lin_hyster_stiff_int_force_v3(VELOCI(:,ITIME),NDOF,KEY(:,ITIME-1),DISPLA(:,ITIME1),DISPLA(:,ITIME),SPRFOR(:,ITIME-1),UT(:,ITIME-1),UC(:,ITIME-1),UO(:,ITIME-1),ITIME,KELAST);
[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f11_nonl
in_hyster_stiff_int_force_cubic(VELOCI(:,ITIME),NDOF,KEY(:,ITIME-1),DISPLA(:,ITIME1),DISPLA(:,ITIME),UT(:,ITIME-1),UC(:,ITIME-1),UO(:,ITIME-1),ITIME,KELAST);

Section 16.6.6.4: Non-linear or linear damping force, internal dashpot force and state
variable of time step i+1

%[FNLDAM(:,ITIME),DASHFO(:,ITIME)]=f12_lin_damp_int_force_v2(NDOF,VELOCI(:,ITIME)-VELOCI(:,ITIME1),DASHFO(:,ITIME-1),CELAST); %Nonlinear damping force and internal dashpot force of time step
i+1
[FNLDAM(:,ITIME),DASHFO(:,ITIME)]=f13_nonlinmet_damp_int_force(NDOF,DISPLA(:,ITIME),VELOCI(:,ITIM
E),CELAST);
end
if EMKEY~=100
return
end

Section 16.6.6.5: Total non-linear force, linear force and pseudo force
FNONLI(:,ITIME)=FNLSTI(:,ITIME)+FNLDAM(:,ITIME); % Nonlinear internal restoring force
FLINEA(:,ITIME)=f14_linear_int_rest_force(K,DISPLA(:,ITIME),Cnp,VELOCI(:,ITIME)); %
Linear internal restoring force vector
PSEUFO(:,ITIME)=FLINEA(:,ITIME)-FNONLI(:,ITIME); % Pseudo force vector

181

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

end

Section 16.6.7: Convergence check


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The convergence criteria is checked for the time steps from
%
CONVTI+1 to LPTISE. If after checking these points some of
%
them have not converged yet, the convergence loop is
%
repeated all over again for those new not converged tme
%
steps. On the other hand, if all the time steps of the
%
current segment (until LPTISE) have converged the
%
convergence loop is finished and a new time segment
%
calculation is started.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
pause(1);
disp(' ')
disp('Convergence check ... ')
OLDCOT=CONVTI;
convlock=1;
while CONVTI~=LPTISE
TT=CONVTI+1;
ST=CONVTI-FPTISE+2;
if norm(PSEUFO(:,TT))==0
PSFERR=norm(PSEUFO(:,TT)-PSFOLD(:,ST));
else
PSFERR=norm(PSEUFO(:,TT)-PSFOLD(:,ST))/norm(PSEUFO(:,TT));
end
if norm(DISPLA(:,TT))==0
DISERR=norm(DISPLA(:,TT)-DISOLD(:,ST));
else
DISERR=norm(DISPLA(:,TT)-DISOLD(:,ST))/norm(DISPLA(:,TT));
end
if PSFERR<=tol && DISERR<=tol
CONVTI=CONVTI+1;
disp('
')
disp('Convergence achieved for the time step number ... ')
disp(CONVTI)
else
PSFOLD(:,OLDCOT-FPTISE+2:NSAMPL)=PSEUFO(:,OLDCOT+1:LPTISE);
DISOLD(:,OLDCOT-FPTISE+2:NSAMPL)=DISPLA(:,OLDCOT+1:LPTISE);
convlock=0;
break
end
end
disp('
')
if convlock==0
NOTCON=CONVTI+1;
disp('New iteration required. Not convergence achieved for the time step number ...
')

182

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

disp(NOTCON)
elseif ITIMSE~=NTIMSE
disp('Convergence achieved for all the time steps within the current segment. Next
segment will be evaluated ... ')
else
disp('Finish: Convergence achieved for all the time steps of the interest span ')
end
end
end

Section 17: Output results post processing


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The output results are postprocessed for each DOF
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
pause(1)
disp('
')
disp('Starting output results post processing... ')
t=LODTIM(1:NRESIT);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Creation of plots tables in excel
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
pause(1)
% Calculation the relative displacement and velocity
DISREL=zeros(NDOF+1,NRESIT);
VELREL=zeros(NDOF+1,NRESIT);
DISREL(1,:)=DISPLA(1,:);
DISREL(NDOF+1,:)=-DISPLA(NDOF,:);
VELREL(1,:)=VELOCI(1,:);
VELREL(NDOF+1,:)=-VELOCI(NDOF,:);
if NDOF>=2
for IDOF=2:NDOF
DISREL(IDOF)=DISPLA(IDOF)-DISPLA(IDOF-1);
VELREL(IDOF)=VELOCI(IDOF)-VELOCI(IDOF-1);
end
end
% Tables and plot generation
nullrow1=zeros(1,NRESIT);
nullrow2=zeros(1,MEXCIT);
for IDOF=1:NDOF+1
disp(' ')
disp('Generating tables and figures of the response of degree of freedom number...')
disp(IDOF)
pause(1)
if IDOF<NDOF+1
f15_excel_tables_hftd_v2(IDOF,name,DISPLA(IDOF,:),VELOCI(IDOF,:),ACCELE(IDOF,:),PSEUFO(IDOF,1:NRE
SIT),FNLSTI(IDOF,:),FNLDAM(IDOF,:),FNONLI(IDOF,:),FLINEA(IDOF,:),KEY(IDOF,:),SPRFOR(IDOF,:),DASHF

183

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

O(IDOF,:),t,UT(IDOF,:),UC(IDOF,:),UO(IDOF,:),EXCITA,DISFRE(IDOF,1:MEXCIT),VELFRE(IDOF,1:MEXCIT),A
CCFRE(IDOF,1:MEXCIT));
f16_plot_resp_hftd_v2(IDOF,name,DISPLA(IDOF,:),VELOCI(IDOF,:),ACCELE(IDOF,:),SPRFOR(IDOF,:),DASHF
O(IDOF,:),DISREL(IDOF,:),VELREL(IDOF,:),t,EXCITA,DISFRE(IDOF,1:MEXCIT),VELFRE(IDOF,1:MEXCIT),ACCF
RE(IDOF,1:MEXCIT),FOUACC(1:MEXCIT));
else
f15_excel_tables_hftd_v2(IDOF,name,nullrow1,nullrow1,nullrow1,nullrow1,nullrow1,nullrow1,nullrow1
,nullrow1,KEY(IDOF,:),SPRFOR(IDOF,:),DASHFO(IDOF,:),t,UT(IDOF,:),UC(IDOF,:),UO(IDOF,:),EXCITA,nul
lrow2,nullrow2,nullrow2);
f16_plot_resp_hftd_v2(IDOF,name,nullrow1,nullrow1,nullrow1,SPRFOR(IDOF,:),DASHFO(IDOF,:),DISREL(I
DOF,:),VELREL(IDOF,:),t,EXCITA,nullrow2,nullrow2,nullrow2,nullrow2);
end
end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Change the name of variables to be saved
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp(' ')
disp('Changing the name of the selected variables to be saved...')
%
The new names of the variables are created and stored using the
%
function "f17_new_name_var"
[vk,vc,vm,vtol,vg,vphi,vnatfreq,vt1,vw1,vxi1,vsignal,vdt,vnsampl,vntimse,vmexcit,vdeltaf,vexcita,
vlopass,vhipass,vnresit,vdispla,vveloci,vaccele,vpseufo,vnlstiff,vnldamp,vfnonli,vflinea,vkey,vsp
rinfor,vdashfor,vdisfre,vvelfre,vaccfre,vt,vut,vuc,vdisrel,vvelrel,vfouacc]=f17_new_name_var_hftd
_v2(name);
%
The new names are assigned to all the variables to be saved.
eval([vk ' = K;']);
eval([vc ' = C;']);
eval([vm ' = M;']);
eval([vtol ' = tol;']);
eval([vg ' = g;']);
eval([vphi ' = Phi;']);
eval([vnatfreq ' = Natfreq;']);
eval([vt1 ' = T1;']);
eval([vw1 ' = w1;']);
eval([vxi1 ' = xi1;']);
eval([vsignal ' = signal;']);
eval([vdt ' = DT;']);
eval([vnsampl ' = NSAMPL;']);
eval([vntimse ' = NTIMSE;']);
eval([vmexcit ' = MEXCIT;']);
eval([vdeltaf ' = DELTAF;']);
eval([vexcita ' = EXCITA;']);
eval([vlopass ' = LOPASS;']);
eval([vhipass ' = HIPASS;']);
eval([vnresit ' = NRESIT;']);
eval([vdispla ' = DISPLA;']);
eval([vveloci ' = VELOCI;']);

184

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

eval([vaccele ' = ACCELE;']);


eval([vpseufo ' = PSEUFO;']);
eval([vnlstiff ' = FNLSTI;']);
eval([vnldamp ' = FNLDAM;']);
eval([vfnonli ' = FNONLI;']);
eval([vflinea ' = FLINEA;']);
eval([vkey ' = KEY;']);
eval([vsprinfor ' = SPRFOR;']);
eval([vdashfor ' = DASHFO;']);
eval([vdisfre ' = DISFRE;']);
eval([vvelfre ' = VELFRE;']);
eval([vaccfre ' = ACCFRE;']);
eval([vt ' = t;']);
eval([vut ' = UT;']);
eval([vuc ' = UC;']);
eval([vdisrel ' = DISREL;']);
eval([vvelrel ' = VELREL;']);
eval([vfouacc ' = FOUACC;']);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The selected variables are saved
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
pause(1)
disp(' ')
disp('Saving selected set of variables...')
%File path where the data is store. The file path can be changed to another
%desired location.
path='C:\Users\slfx\Documents\Fernando\Graduation project\Master thesis\Analysis files\Matlab
models\HFTD_freq dependence\HFTD Output results\Variables\';
%path='C:\Users\Fernando Sirumbal\Documents\Documents FERNANDO\TU Delft\Academic\MSc
thesis\Master thesis\Analysis files\Matlab models\HFTD_freq dependence\HFTD Output
results\Variables\';
%Name of the matlab data files
filename1=[path 'results_' name]; %Name of the matlab file containing the results variables
filename2=[path 'data_' name]; %Name of the matlab file conatining the data variables
%Saving the variables in both files
save(filename1,vdispla,vveloci,vaccele,vpseufo,vnlstiff,vnldamp,vfnonli,vflinea,vkey,vsprinfor,vd
ashfor,vdisfre,vvelfre,vaccfre,vt,vfouacc,vut,vuc,vdisrel,vvelrel);
save(filename2,vk,vc,vm,vtol,vg,vphi,vnatfreq,vt1,vw1,vxi1,vsignal,vdt,vnsampl,vntimse,vmexcit,vd
eltaf,vexcita,vlopass,vhipass,vnresit);
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Delete the variables from the workspace
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp(' ')
disp('Deleting all variables from the workspace...')
clear;
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
End of the program
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

185

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

disp(' ')
disp('End of the HFTD program... Check the results files saved in the specified folder')

Published with MATLAB R2012b

186

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

B.1.2. HFTD selected functions

function [DCYFAC]=f5_decay_func(INPFAC,LPTISE,NDCAYF,DT,TDCAYF)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to determine the factors of the decaying function appended to
%
the ground acceleration and pseudo force.
%
%
inputs
%
%
INPFAC Vector containing the input (ground acceleration or pseudo
%
force)factors for each time step.
%
LPTISE Index indicating the last time point of the current
%
segment.
%
NDCAYF Number of time steps defining the length of the decaying
%
function.
%
DT
Time step size.
%
TDCAYF Decaying function length.
%
%
outputs
%
%
DCYFAC Decaying function factors for the input (ground
%
acceleration or pseudo force). Different options for
%
decaying function expressions can be used by uncomment the
%
chosen one.
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
f0=INPFAC(LPTISE);
DCYFAC=zeros(1,NDCAYF);
for k=1:NDCAYF
DCYFAC(k)=f0*(1-k*DT/TDCAYF);
%DCYFAC(k)=f0*(1-k*DT/TDCAYF+sin(2*pi*k*DT/TDCAYF)/(2*pi));
end

Published with MATLAB R2012b

187

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

function [FOUREA,FOUIMA]=f6_fast_fourier_transf(x,N,MEXCIT)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to apply the direct discrete Fourier transform to a time
%
dependent variable
%
%
inputs
%
%
x
Time dependent variable (signal) to which direct Fourier
%
tranform is applied.
%
N
Number of time points contained in x.
%
MEXCIT Maximum number of frequency samples to be analyzed.
%
%
%
outputs
%
%
FOUREA Real part of the direct Fourier transform of signal x.
%
FOUIMA Imaginary part of the direct Fourier transform of signal x.
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% FFT computation
y = fft(x,N);
rey=real(y);
imy=imag(y);
% Scale FFT output
reh=rey/N;
imh=imy/N;
% Copy only the frequency responses that are going to be used in the
% frequency domain analysis. The conjugates are not required to be copied
% since they do not provide additional information.
FOUREA(1,:)=reh(1:MEXCIT);
FOUIMA(1,:)=imh(1:MEXCIT);

Published with MATLAB R2012b

188

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

function
[DIFREA,DIFRIM]=f7_solver_pseudo_lin_direct_freq_v2(K,C,Cfd,Chy,M,w,NDOF,FOUACR,FOUACI,FOUPFR,FOU
PFI)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to solve a linear system in the frequency domain using the
%
direct method
%
%
[-(w^2)*M + i*w*C + K] u(w) = -M*I*ug(w) + q(w)
%
%
inputs
%
%
K
Linear elastic stiffness matrix of the system.
%
C
Linear elastic constant damping matrix of the system.
%
Cfd
Linear elastic frequency dependent damping matrix of the
%
system.
%
Chy
Linear elastic hysteretic damping matrix of the system.
%
M
Mass matrix of the system.
%
w
Excitation frequency.
%
NDOF
Number of degres of freedom.
%
FOUACR Fourier transform of the ground acceleration (real part)
%
corresponding to the excitation frequency w
%
FOUACI Fourier transform of the ground acceleration (imaginary part)
%
corresponding to the excitation frequency w
%
FOUPFR Fourier transform of the pseudo force vector (real part)
%
corresponding to the excitation frequency w
%
FOUPFI Fourier transform of the pseudo force vector (imaginary part)
%
corresponding to the excitation frequency w
%
%
outputs
%
%
DIFREA Displacement vector in the frequency domain (real part)
%
obtained from the solution of the linear system of
%
equations.
%
%
DIFRIM Displacement vector in the frequency domain (imaginary
%
part) obtained from the solution of the linear system of
%
equations.
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Components of the modified stiffness matrix in the frequency domain
A=K-(w^2)*M;
B=w*(C+Cfd+Chy);
% Assembling the modified stiffness matrix

189

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

KMODF=[A B;-B A];


% Rigig body motion vector
l=ones(NDOF,1);
% Real part of the modified force vector
FMODRE=(-FOUACR*M*l)+FOUPFR;
% Imaginary part of the modified force vector
FMODIM=(-FOUACI*M*l)+FOUPFI;
% Assembling the modified force vector
FORMDF=[FMODIM;FMODRE];
% Solution of the system of equations
SOL=KMODF\FORMDF;
% Spliting the solution between real and imaginary terms
DIFRIM=SOL(1:NDOF);
DIFREA=SOL(NDOF+1:2*NDOF);

Published with MATLAB R2012b

190

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

function [IFOURE,IFOUIM]=f8_inv_fast_fourier_transf(x,N,FPTISE,LPTISE)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to apply the inverse discrete Fourier transform to a time
%
dependent variable
%
%
inputs
%
%
x
Frequency dependent complex variable (spectrum) to which inverse Fourier
%
tranform is applied.
%
N
Number of frequency points contained in x.
%
FPTISE First time point of the signal that is copied as output
%
LPTISE Last time point of the signal that is copied as output
%
%
outputs
%
%
IFOURE Real part of the inverse Fourier transform of spectrum x.
%
IFOUIM Imaginary part of the inverse Fourier transform of spectrum x.
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% IFFT computation
y = ifft(N*x,N,'symmetric');
rey=real(y);
imy=imag(y);
% Copy only the results of the time steps within the current time segment
IFOURE=rey(FPTISE:LPTISE);
IFOUIM=imy(FPTISE:LPTISE);

Published with MATLAB R2012b

191

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

function FLINEA=f14_linear_int_rest_force(K,DISPLA,C,VELOCI)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal linear restoring force
%
%
%
inputs
%
%
K
Stiffness matrix of the system
%
DISPLA Displacement vector
%
C
Damping matrix of the system
%
VELOCI Velocity vector
%
%
%
outputs
%
%
FLINEA Vector containing the linear restoring force
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
FLINEA=K*DISPLA+C*VELOCI;

Published with MATLAB R2012b

192

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

B.2. Newmark method code

B.2.1. Main Newmark script

Section 1: File name .................................................................................................................... 194


Section 2: Newmark method parameters ....................................................................................... 194
Section 3: Exercise data definition ................................................................................................ 195
Section 4: Construction of the linear elastic matrices of the system................................................. 198
Section 5: Free vibration Eigen-analysis ........................................................................................ 199
Section 6: Ground acceleration signal information .......................................................................... 200
Section 7: Time step size ............................................................................................................. 201
Section 8: Time span of interest ................................................................................................... 202
Section 9: Ground acceleration signal modification ......................................................................... 203
Section 10: Initialization of variables ............................................................................................. 203
Section 11: Newmark method execution for each time step ............................................................ 204
Section 12: Output results post processing .................................................................................... 205

193

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Newmark method for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
This script execute the Newmark method for linear or nonlinear simplfied
%
dynamic models of one or more degrees of freedom (DOF's) subjected to
%
base (ground) acceleration.
%
%
After defining the system properties and the load signal, the following
%
results are obtained: time history displacement, velocity, acceleration
%
and nonlinear internal forces of the system.
%
%
The name of the function indicating the type of nonlinearity to be used
%
in the problem is indicated in Section 11. The functions describing
%
type of tangential stiffnes and damping matrix should be selected.
%
Besides, the non-linear stiffness and damping models are defined inside
%
the function "f18_Newt_Raph_modf_v6".
%
%
The output results are saved as excel tables, figures and matlab data
%
files. The file path where these output results files are saved should
%
be indicated Section 12 of this script and in the function files
%
"f19_excel_tables_newm" and "f20_plot_resp_newm".
%
%

Section 1: File name


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
All the output results of the program will be saved in the folder
%
"Output results" located in the work folder. The name specified by
%
the user is used to save the output files.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
clear;
disp('==========================================================================')
disp('
Problem data
')
disp('==========================================================================')
name=input('\nEnter the name of the file (this name will be used to save the output data)...
','s');
disp('
')
pause(2);

Section 2: Newmark method parameters


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%

The Newmark parameters gamma and beta should be selected. Average

194

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%
and linear acceleration cases are the two default options.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Define the Newmark method to be executed
')
disp('==========================================================================')
lock=0;
while lock==0
pause(2);
opt=input('\nSelect a Newmark special case or the numerical values of the parameters gamma
and beta. \nEnter "AA" for Average Acceleration or enter "LA" for Linear Acceleration, \nor "OC"
for other case .... ','s');
h=strcmp(opt,{'AA','LA','OC'});
if h(1)==1
gama=0.5;
beta=0.25;
lock=1;
elseif h(2)==1
gama=0.5;
beta=1/6;
lock=1;
elseif h(3)==1
pause(1);
gama=input('\nEnter the value of gamma) .... ');
if gama<0 || gama>1
disp('
')
disp('Error message: Not valid specified value for gamma')
else
pause(1);
beta=input('\nEnter the value of beta) .... ');
if beta<0 || beta>1
disp('
')
disp('Error message: Not valid specified value for beta')
else
lock=1;
end
end
else
pause(2);
disp('
')
disp('Error message: Option not correctly chosen. Choose one of the three available
options')
end
end

Section 3: Exercise data definition


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Input the mass, damping and stiffness matrices of the simplified MDOF

195

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%
system. The gravity constant and the tolerance and maximum number of
%
iterations for non-linear analysis are also input by the user.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Define the elastic stiffness, elastic damping and mass matrices
')
disp('==========================================================================')
pause(1);
lock=0;
while lock==0
MCOF=input('\nEnter the masses of each degree of freedom m_i \n(in a row vector enter one
mass per DOF).... ');
sizem=size(MCOF);
if sizem(1)~=1
disp(' ')
disp('Error message: The masses should be arranged in a row vector')
pause(2);
else
lock=1;
end
end
NDOF=sizem(2);
pause(1);
disp(' ')
disp('The number of degrees of freedom (NDOF) of the system is defined to be ')
NDOF
pause(1)
disp(' ')
disp('The masses of each degree of freedom are defined to be ')
MCOF
pause(1);
lock=0;
while lock==0
KELAST=input('\nEnter the linear elastic stiffness coefficients k_i of the springs \n(in a
row vector enter NDOF+1 stiffness coefficients) .... ');
sizek=size(KELAST);
if sizek(1)~=1
disp('
')
disp('Error message: The spring coefficients should be arranged in a row vector')
pause(2);
elseif sizek(2)~=NDOF+1
disp(' ')
disp('Error message: The row vectors defining the coefficientes of K should have NDOF+1
components ')
pause(2);
else
lock=1;
end
end
disp(' ')

196

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

disp('The linear elastic stiffness coefficients are defined to be')


KELAST
pause(1);
lock=0;
while lock==0
CELAST=input('\nEnter the linear elastic damping coefficients c_i of the dashpots \n(in a row
vector enter NDOF+1 damping coefficients)... ');
sizec=size(CELAST);
if sizec(1)~=1
disp(' ')
disp('Error message: The dashpots coefficients should be arranged in a row vector')
pause(2);
elseif sizec~=NDOF+1
disp(' ')
disp('Error message: The row vectors defining the coefficientes of C should have NDOF+1
components ')
pause(2);
else
lock=1;
end
end
disp(' ')
disp('The linear elastic damping coefficients are defined to be ')
CELAST
pause(1);
lock=0;
while lock==0
g=input('\nEnter the value of the gravity constant g .... ');
sizeg=size(g);
if sizeg(1)~= 1 || sizeg(2)~=1
disp(' ')
disp('Error message: g must be a scalar value')
pause(2);
elseif g<=0
disp(' ')
disp('Error message: g must be defined as a positive value')
pause(2);
else
lock=1;
end
end
disp(' ')
disp('The gravity constant is defined to be ')
g
disp('
')
pause(1);
lock=0;
while lock==0
tol=input('\nEnter the convergence tolerance tol for noninear analysis.... ');
sizet=size(tol);

197

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

if sizet(1)~= 1
disp(' ')
disp('Error
pause(2);
elseif tol>0.01
disp(' ')
disp('Error
pause(2);
else
lock=1;
end

Appendix B: MATLAB Routines

|| sizet(2)~=1
message: tol must be a scalar value')
|| tol<=0
message: tol must be positive and smaller than 10^-2')

end
disp(' ')
disp('The tolerance for convergence check is defined to be ')
tol
pause(1);
lock=0;
while lock==0
maxite=input('\nEnter the maximum number of iterations for convergence check in nonlinear
analysis .... ');
sizemax=size(maxite);
if sizemax(1)~= 1 || sizemax(2)~=1
disp(' ')
disp('Error message: maxite must be a scalar value')
pause(2);
elseif mod(maxite,1)~=0 || maxite<=0
disp(' ')
disp('Error message: maxite must be positive and integer')
pause(2);
else
lock=1;
end
end
disp(' ')
disp('The maximum number of iterations for convergence check is defined to be ')
maxite
pause(2);

Section 4: Construction of the linear elastic matrices of the system


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Based on the coefficients values defined in the previous section for
%
each mass, spring and dashpot, the linear elastic matrices of the
%
system are constructed using the function file "f1_linelast_mat".
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Linear elastic matrices
')
disp('==========================================================================')

198

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%Mass matrix
M=zeros(NDOF);
for IDOF=1:NDOF
M(IDOF,IDOF)=MCOF(IDOF);
end
l=ones(NDOF,1);
P=-M*l;
disp(' ')
disp('The mass matrix M is')
M
pause(1);
%Stiffness matrix
K=f1_linelast_mat(NDOF,KELAST);
disp(' ')
disp('The linear elastic stiffness matrix K is')
K
pause(1);
%Damping matrix
C=f1_linelast_mat(NDOF,CELAST);
disp(' ')
disp('The linear elastic damping matrix C is')
C
pause(2);

Section 5: Free vibration Eigen-analysis


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Eigen-analysis of the simplified system. Mode shape vectors,
%
eigen-frequencies, damping ratio and generalized matrices of the system
%
are calculated.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Free vibration natural frequencies and mode shape vectors
disp('==========================================================================')
[Phi,Natfreq]=eig(K,M);
Mgen=transpose(Phi)*M*Phi;
Kgen=transpose(Phi)*K*Phi;
Cgen=transpose(Phi)*C*Phi;
w1=sqrt(Natfreq(1,1));
T1=2*pi/w1;
f1=1/T1;
xi1=(1/(2*w1))*(Cgen(1,1)/Mgen(1,1));
pause(2);
disp('
')
disp('The eigenvectors matrix is ')
disp('
')
Phi
pause(1);

')

199

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

disp('
')
disp('The eigenvalues (squared natural circular frequencies) matrix is')
disp('
')
Natfreq
pause(1);
disp('
')
disp('The fundamental period of the system is defined to be')
disp('
')
T1
pause(1);
disp('
')
disp('The fundamental circular frequency of the system is defined to be')
disp('
')
w1
pause(1);
disp('
')
disp('The damping ratio of the fundamental mode of vibration of the system is defined to be')
disp('
')
xi1
disp('
')
pause(2);

Section 6: Ground acceleration signal information


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The data of the ground acceleration signal must be stored as a
%
matlab variable called "signal". This variable should be
%
defined as a matrix with two columns and as many rows as time
%
points available in the signal data. The first column contains the
%
time steps values, which should start in time = 0 seconds, and
%
should be equally spaced (constant time step size). The second
%
colum contains the ground acceleration values normalized with the
%
gravity acceleration g.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Ground acceleration signal information
')
disp('==========================================================================')
disp('
')
disp('Loading the ground acceleration variable "signal" from file signal.mat saved in the work
folder...')
disp('
')
pause(2);
load signal;
lock=1;
while lock~=0
sizesig=size(signal);
[SIGTIM,NFACT,SIGNDT,lock,k]=f3_ground_accel_signal_check(signal,sizesig);% Function that
checks the correct definition of the signal variable

200

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

if lock==1
disp('Error message:
return
elseif lock==2
disp('Error message:
return
elseif lock==3
disp('Error message:
disp('
')
pause(2);
fprintf('The problem
return
end

Appendix B: MATLAB Routines

The first time step must be zero')

Signal data variable must have two columns')

Time step size of signal data is not constant')

was found between the time points %d and %d\n',k,k+1)

end
SIGFAC=g*signal(:,2);
TACSIG=NFACT*SIGNDT;
pause(2);
disp('
')
disp('The number of time points contained in the ground acceleration signal is ')
disp('
')
NFACT
pause(1);
disp('
')
disp('The time step size of the ground acceleration signal is ')
disp('
')
SIGNDT
pause(1);
disp('
')
disp('The duration of the ground acceleration signal is ')
disp('
')
TACSIG
pause(1);
disp('
')
disp('The maximum ground acceleration magnitude contained in the signal is ')
disp('
')
disp(max(abs(SIGFAC)))
disp('
')
pause(2);

Section 7: Time step size


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The time step size (DT) is calculated base on the signal
%
discretization.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Time step size
')
disp('==========================================================================')

201

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

lock=0;
while lock==0
pause(2);
opt=input('\nSelect the option for the definition of the HFTD time step size (DT). Enter "0"
for DT to be equal to the signal step \nsize (SIGNDT) or enter "1" for DT to be a multiple of
SIGNDT .... ');
if opt==0
TIMES=1;
lock=1;
elseif opt==1
pause(2);
TIMES=input('\nEnter the number of times that SIGNDT should be multiplied to obtain (DT)
.... ');
if mod(TIMES,1)~=0 || TIMES<=0
disp('
')
disp('Error message: The input value should be a positive integer')
else
lock=1;
end
else
pause(2);
disp('
')
disp('Error message: The input value must be 0 or 1')
end
end
DT=TIMES*SIGNDT;
pause(1);
disp('
')
disp('The Newmark time step size (DT) is defined to be')
disp('
')
DT
pause(2)

Section 8: Time span of interest


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Time span for which the reponses are calculated.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('==========================================================================')
disp('
Time span of interest
')
disp('==========================================================================')
lock=0;
while lock==0
pause(1);
TRESP=input('\nEnter the time span of interest in seconds .... ');
NFRESP=ceil(TRESP/DT);
TRESP=DT*NFRESP;
if TRESP<=0
disp('
')

202

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

disp('Error message: The time span of interest should be positive')


else
lock=1;
end
end
pause(1);
disp('
disp('The
be')
disp('
NFRESP
pause(1);
disp('
disp('The
disp('
TRESP
disp('
pause(2);

')
number of time steps defining the time span of interest length (NFRESP) is defined to
')

')
time span of interest for determining the response (TRESP) is defined to be')
')
')

Section 9: Ground acceleration signal modification


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The ground acceleration data is cutted or extended adding zeros
%
until the response time span is covered.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LODTIM=zeros(1,NFRESP);
LODFAC=zeros(1,NFRESP);
for k=1:NFRESP
LODTIM(k)=(k-1)*DT;
j=1+(k-1)*TIMES;
if j<=NFACT
LODFAC(k)=SIGFAC(j);
else
LODFAC(k)=0;
end
end

Section 10: Initialization of variables


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Initialize response and internal restoring force variables
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
DISPLA=zeros(NDOF,NFRESP);
VELOCI=zeros(NDOF,NFRESP);
ACCELE=zeros(NDOF,NFRESP);
FNLSTI=zeros(NDOF,NFRESP);
SPRFOR=zeros(NDOF+1,NFRESP);
FNLDAM=zeros(NDOF,NFRESP);
DASHFO=zeros(NDOF+1,NFRESP);

203

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

KEYSTI=zeros(NDOF+1,NFRESP);
KEYDAM=zeros(NDOF+1,NFRESP);
UT=zeros(NDOF+1,NFRESP);
UC=zeros(NDOF+1,NFRESP);
UO=zeros(NDOF+1,NFRESP);
NITERA=zeros(1,NFRESP);

Section 11: Newmark method execution for each time step


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The Newmark calculation process is executed step by step. A time
%
step loop is defined in order to repeat the process for the
%
NFRESP time steps previously determined.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('==========================================================================')
disp('
Starting Newmark method execution for each time step
')
disp('==========================================================================')
ACCELE(:,1)=M\(LODFAC(1)*P-FNLDAM(:,1)-FNLSTI(:,1)); % Initial acceleration
for ITIME=1:NFRESP-1
VARLOD=(LODFAC(ITIME+1)-LODFAC(ITIME))*P; %Load increment
%
%Tangent stiffness matrix
%STIMAT=f15_tang_stiff_mat_elastoplast_hyster(NDOF,KEYSTI(:,ITIME),KELAST);
STIMAT=f16_tang_stiff_mat_cubic_hyster_v2(NDOF,KEYSTI(:,ITIME),KELAST,DISPLA(:,ITIME),UO(:,ITIME)
);
%STIMAT=K;
%
%Tangent damping matrix
%DAMMAT=C
DAMMAT=f17_tang_damp_mat_nonlimet(NDOF,KEYDAM(:,ITIME),CELAST,DISPLA(:,ITIME));
%
EFFSTI=STIMAT+((gama/beta/DT)*DAMMAT)+((1/beta/(DT^2))*M); % Effective stiffness matrix
EFVALO=VARLOD+(((1/beta/DT)*M+(gama/beta)*DAMMAT)*VELOCI(:,ITIME))+(((1/2/beta)*M+((gama/2/beta)1)*DT*DAMMAT)*ACCELE(:,ITIME)); %Effective load increment
[DISPLA(:,ITIME+1),VELOCI(:,ITIME+1),ACCELE(:,ITIME+1),FNLSTI(:,ITIME+1),FNLDAM(:,ITIME+1),NITERA
(ITIME),KEYSTI(:,ITIME+1),SPRFOR(:,ITIME+1),DASHFO(:,ITIME+1),UT(:,ITIME+1),UC(:,ITIME+1),UO(:,IT
IME+1),EMKEY]=f18_Newt_Raph_modf_v6(NDOF,DISPLA(:,ITIME),VELOCI(:,ITIME),ACCELE(:,ITIME),FNLSTI(:
,ITIME),FNLDAM(:,ITIME),EFVALO,EFFSTI,M,tol,KEYSTI(:,ITIME),SPRFOR(:,ITIME),DASHFO(:,ITIME),UT(:,
ITIME),UC(:,ITIME),UO(:,ITIME),ITIME,maxite,gama,beta,DT,LODFAC(ITIME+1),P,VARLOD,KELAST,CELAST);
if EMKEY~=100
return
end
if NITERA(ITIME)>maxite
return
else
fprintf('\nConvergence achieved for time step number %d',ITIME)

204

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

disp('

Appendix B: MATLAB Routines

')

end
end
pause(2)

Section 12: Output results post processing


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The output results are postprocessed for each DOF
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp('
')
disp('Starting output results post processing... ')
t=LODTIM;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Creation of plots and tables in excel
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Calculation of the relative displacement and velocity
DISREL=zeros(NDOF+1,NFRESP);
VELREL=zeros(NDOF+1,NFRESP);
DISREL(1,:)=DISPLA(1,:);
DISREL(NDOF+1,:)=-DISPLA(NDOF,:);
VELREL(1,:)=VELOCI(1,:);
VELREL(NDOF+1,:)=-VELOCI(NDOF,:);
if NDOF>=2
for IDOF=2:NDOF
DISREL(IDOF)=DISPLA(IDOF)-DISPLA(IDOF-1);
VELREL(IDOF)=VELOCI(IDOF)-VELOCI(IDOF-1);
end
end
% Tables and plots generation
nullrow=zeros(1,NFRESP);
for IDOF=1:NDOF+1
disp(' ')
disp('Generating tables and figures of the response of degree of freedom number...')
disp(IDOF)
pause(1)
if IDOF<NDOF+1
f19_excel_tables_newm(IDOF,name,DISPLA(IDOF,:),VELOCI(IDOF,:),ACCELE(IDOF,:),FNLSTI(IDOF,:),FNLDA
M(IDOF,:),KEYSTI(IDOF,:),SPRFOR(IDOF,:),DASHFO(IDOF,:),t,UT(IDOF,:),UC(IDOF,:),UO(IDOF,:))
f20_plot_resp_newm(IDOF,name,DISPLA(IDOF,:),VELOCI(IDOF,:),ACCELE(IDOF,:),SPRFOR(IDOF,:),DASHFO(I
DOF,:),DISREL(IDOF,:),VELREL(IDOF,:),t)
else
f19_excel_tables_newm(IDOF,name,nullrow,nullrow,nullrow,nullrow,nullrow,KEYSTI(IDOF,:),SPRFOR(IDO
F,:),DASHFO(IDOF,:),t,UT(IDOF,:),UC(IDOF,:),UO(IDOF,:))
f20_plot_resp_newm(IDOF,name,nullrow,nullrow,nullrow,SPRFOR(IDOF,:),DASHFO(IDOF,:),DISREL(IDOF,:)

205

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

,VELREL(IDOF,:),t)
end
end
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Changing the name of variables to be saved
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp(' ')
disp('Changing the name of the selected variables to be saved...')
%
The new names of the variables are created and stored using the
%
function f21_new_name_var_newm
[vk,vc,vm,vtol,vg,vphi,vnatfreq,vt1,vw1,vxi1,vsignal,vdt,vdispla,vveloci,vaccele,vnlstiff,vnldamp
,vkeystiff,vkeydamp,vsprinfor,vdashfor,vt,vnitera,vut,vuc,vdisrel,vvelrel]=new_name_var_newm(name
);
%
The new names are assigned to all the variables to be saved.
eval([vk ' = K;']);
eval([vc ' = C;']);
eval([vm ' = M;']);
eval([vtol ' = tol;']);
eval([vg ' = g;']);
eval([vphi ' = Phi;']);
eval([vnatfreq ' = Natfreq;']);
eval([vt1 ' = T1;']);
eval([vw1 ' = w1;']);
eval([vxi1 ' = xi1;']);
eval([vsignal ' = signal;']);
eval([vdt ' = DT;']);
eval([vdispla ' = DISPLA;']);
eval([vveloci ' = VELOCI;']);
eval([vaccele ' = ACCELE;']);
eval([vnlstiff ' = FNLSTI;']);
eval([vnldamp ' = FNLDAM;']);
eval([vkeystiff ' = KEYSTI;']);
eval([vkeydamp ' = KEYDAM;']);
eval([vsprinfor ' = SPRFOR;']);
eval([vdashfor ' = DASHFO;']);
eval([vt ' = t;']);
eval([vnitera ' = NITERA;']);
eval([vut ' = UT;']);
eval([vuc ' = UC;']);
eval([vdisrel ' = DISREL;']);
eval([vvelrel ' = VELREL;']);
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The selected variables are saved
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
pause(1)
disp(' ')
disp('Saving selected set of variables...')
%File path where the data is store. The file path can be changed to another

206

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%desired location.
path='C:\Users\slfx\Documents\Fernando\Graduation project\Master thesis\Analysis files\Matlab
models\Newmark_nonli_MDOF\Newm Output results\';
%path='C:\Users\Fernando Sirumbal\Documents\Documents FERNANDO\TU Delft\Academic\MSc
thesis\Master thesis\Analysis files\Matlab models\Newmark_nonli_MDOF\Newm Output results\';
%Name of the matlab data files
filename1=[path 'results_' name]; %Name of the matlab file containing the results variables
filename2=[path 'data_' name]; %Name of the matlab file conatining the data variables
%Saving the variables in both files
save(filename1,vdispla,vveloci,vaccele,vkeystiff,vkeydamp,vsprinfor,vdashfor,vt,vnlstiff,vnldamp,
vnitera,vut,vuc,vdisrel,vvelrel);
save(filename2,vk,vc,vm,vtol,vg,vphi,vnatfreq,vt1,vw1,vxi1,vsignal,vdt);
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Deleting all the variables
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
pause(1)
disp(' ')
disp('Deleting all variables from the workspace...')
clear;
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
End of the program
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
pause(1)
disp(' ')
disp('End of Newmark program... Check the results files saved in the specified folder')

Published with MATLAB R2012b

207

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

B.2.2. Newmark selected functions

function STIMAT=f15_tang_stiff_mat_elastoplast_hyster(NDOF,KEY,KELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Newmark method for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the nonlinear tangential stiffness matrix of the
%
current time step for the linear elastoplastic model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system
%
KEY
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
%
%
outputs
%
%
STIMAT
Tangential stiffness matrix of the current time step
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
F=zeros(NDOF+1,NDOF);
STIMAT=zeros(NDOF);
if KEY(1)==0
F(1,1)=KELAST(1);
end
if KEY(NDOF+1)==0
F(NDOF+1,NDOF)=-KELAST(NDOF+1);
end
if NDOF>1
for IDOF=2:NDOF
if KEY(IDOF)==0
F(IDOF,IDOF-1)=-KELAST(IDOF);
F(IDOF,IDOF)=KELAST(IDOF);
end
end
end

208

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

for IDOF=1:NDOF
STIMAT(IDOF,:)=F(IDOF,:)-F(IDOF+1,:);
end

Published with MATLAB R2012b

209

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

function STIMAT=f16_tang_stiff_mat_cubic_hyster_v2(NDOF,KEY,KELAST,DISPLA,UO)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Newmark method for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the nonlinear tangential stiffness matrix of the
%
current time step for the cubic elastoplastic model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system.
%
KEY
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
DISPLA
Displacement vector of the previous time step
%
UO
Zero stress displacement of the previous time step
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
%
%
outputs
%
%
STIMAT
Tangential stiffness matrix of the current time step.
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
%Relative displacement
DISREL=zeros(NDOF+1,1);
DISREL(1)=DISPLA(1);
DISREL(NDOF+1)=-DISPLA(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL(IDOF)=DISPLA(IDOF)-DISPLA(IDOF-1);
end
end
% Stiffness matrix coefficients
KTANG=8-(32/9)*((DISREL-UO).*(DISREL-UO));
F=zeros(NDOF+1,NDOF);
STIMAT=zeros(NDOF);
if KELAST(1)==0
F(1,1)=0;
elseif KEY(1)==0 || KEY(1)==3
F(1,1)=KTANG(1);
elseif KEY(1)==2

210

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

F(1,1)=KELAST(1);
end
if KELAST(NDOF+1)==0
F(NDOF+1,NDOF)=0;
elseif KEY(NDOF+1)==0 || KEY(NDOF+1)==3
F(NDOF+1,NDOF)=-KTANG(NDOF+1);
elseif KEY(NDOF+1)==2
F(NDOF+1,NDOF)=-KELAST(NDOF+1);
end
if NDOF>1
for IDOF=2:NDOF
if KELAST(IDOF)==0
F(IDOF,IDOF-1)=0;
F(IDOF,IDOF)=0;
elseif KEY(IDOF)==0 || KEY(IDOF)==3
F(IDOF,IDOF-1)=-KTANG(IDOF);
F(IDOF,IDOF)=KTANG(IDOF);
elseif KEY(IDOF)==2
F(IDOF,IDOF-1)=-KELAST(IDOF);
F(IDOF,IDOF)=KELAST(IDOF);
end
end
end
for IDOF=1:NDOF
STIMAT(IDOF,:)=F(IDOF,:)-F(IDOF+1,:);
end

Published with MATLAB R2012b

211

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

function DAMMAT=F17_tang_damp_mat_nonlimet(NDOF,KEY,CELAST,DISPLA)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Newmark method for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the nonlinear tangential damping matrix of the
%
current time step for the quadractic displacement dependent viscous
%
model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system.
%
KEY
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
DISPLA
Displacement vector of the previous time step
%
UO
Zero stress displacement of the previous time step
%
CELAST
Vector containing the elastic damping coefficients of
%
the dashpots
%
%
%
outputs
%
%
DAMMAT
Tangential damping matrix of the current time step.
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
%Nonlinear damping parameters
alfa=1;
a=zeros(NDOF+1,1);
for IDOF=1:NDOF+1
if CELAST(IDOF)>0
a(IDOF)=alfa;
elseif CELAST(IDOF)<0
a(IDOF)=-alfa;
end
end
%Relative displacement
DISREL=zeros(NDOF+1,1);
DISREL(1)=DISPLA(1);
DISREL(NDOF+1)=-DISPLA(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL(IDOF)=DISPLA(IDOF)-DISPLA(IDOF-1);

212

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

end
end
% Damping matrix coefficients
F=zeros(NDOF+1,NDOF);
DAMMAT=zeros(NDOF);
CTANG=transp(CELAST).*(1+a.*(DISREL.^2));
if KEY(1)==0
F(1,1)=CTANG(1);
end
if KEY(NDOF+1)==0
F(NDOF+1,NDOF)=-CTANG(NDOF+1);
end
if NDOF>1
for IDOF=2:NDOF
if KEY(IDOF)==0
F(IDOF,IDOF-1)=-CTANG(IDOF);
F(IDOF,IDOF)=CTANG(IDOF);
end
end
end
for IDOF=1:NDOF
DAMMAT(IDOF,:)=F(IDOF,:)-F(IDOF+1,:);
end

Published with MATLAB R2012b

213

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

function
[DISPLAOUT,VELOCOUT,ACCELEOUT,FNLSTIOUT,FNLDAMOUT,NITERA,KEYSTIOUT,SPRFOROUT,DASHFOUT,UTOUT,UCOUT
,UOOUT,EMKEY]=f18_Newt_Raph_modf_v6(NDOF,DISPLAIN,VELOCIN,ACCELEIN,FNLSTIN,FNLDAMIN,EFVALO,EFFSTI
,M,tol,KEYSTIN,SPRFORIN,DASHFOIN,UTIN,UCIN,UOIN,ITIME,maxite,gama,beta,DT,LODFAC,P,VARLOD,KELAST,
CELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Newmark method for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Modified Newton Raphson iterative procedure to calculate the non-linear
%
response of the system. The file containing the non-linear models for
%
stiffness and damping should be selected below
%
%
inputs
%
%
%
NDOF
Number of degrees of freedom of the system.
%
DISPLAIN
Displacement vector of the previous time step
%
VELOCIN
Velocity vector of the previous time step
%
ACCELEIN
Acceleration vector of the previous time step
%
FNLSTIN
Non-linear restoring stiffness force vector of the
%
previous time step
%
FNLDAMIN
Non-linear restoring damping force vector of the
%
previous time step
%
EFVALO
Effective load increment of the current time step
%
EFFSTI
Effective stiffness matrix of the current time step
%
M
Mass matrix of the system
%
tol
Tolerance for convergence criteria
%
KEYSTIN
Vector specifying the stiffness nonlinear state of each
%
degree of freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
SPRFORIN
Spring internal stiffness force of the previous time
%
step
%
DASHFOIN
Dashpot internal damping force of the previous time
%
step
%
UTIN
Maximum tensile displacement in the elastic range of
%
the previous time step
%
UCIN
Maximum compression displacement in the elastic range
%
of the previous time step
%
UOIN
Zero stress displacement of the previous time step
%
ITIME
Current time step index
%
maxite
Maximum number of iterations
%
gamma
gamma parameter of the Newmark method
%
beta
beta parameter of the Newmark method
%
DT
Time step size

214

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%
LODFAC
Load factor of the current time step
%
P
Load vector
%
VARLOD
Load increment of the current time step
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
CELAST
Vector containing the elastic damping coefficients of
%
the dashpots
%
%
%
outputs
%
%
DISPLAOUT
Displacement vector of the current time step
%
VELOCOUT
Velocity vector of the current time step
%
ACCELEOUT
Acceleration vector of the current time step
%
FNLSTIOUT
Non-linear restoring stiffness force vector of the
%
current time step
%
FNLDAMOUT
Non-linear restoring damping force vector of the
%
current time step
%
NITERA
Number of iterations required to achieve convergence
%
KEYSTIOUT
Vector specifying the stiffness nonlinear state of each
%
degree of freedom in the current time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
SPRFOROUT
Spring internal stiffness force of the current time
%
step
%
DASHFOUT
Dashpot internal damping force of the current time
%
step
%
UTOUT
Maximum tensile displacement in the elastic range of
%
the current time step
%
UCOUT
Maximum compression displacement in the elastic range
%
of the current time step
%
UOOUT
Zero stress displacement of the current time step
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Initialization of variables
u=zeros(NDOF,maxite+1);
fs=zeros(NDOF,maxite+1);
fd=zeros(NDOF,maxite+1);
dr=zeros(NDOF,maxite+1);
du=zeros(NDOF,maxite+1);
u(:,1)=DISPLAIN;
fs(:,1)=FNLSTIN;
fd(:,1)=FNLDAMIN;
dr(:,1)=EFVALO;
kt=EFFSTI;
error=1;
j=0;

215

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%Start of iterative procedure


while error>tol
j=j+1;
if j==maxite+1
NITERA=j;
fprintf('Error message: Maximum number of iterations (maxite) equal to %d has been
reached without convergence\nin the time step number %d',maxite,ITIME+1)
return
end
du(:,j)=kt\dr(:,j);
u(:,j+1)=u(:,j)+du(:,j);
VARDIS=u(:,j+1)-DISPLAIN; %Displacement increment.
VARVEL=((gama/beta/DT)*VARDIS)-((gama/beta)*VELOCIN)+((1-gama/2/beta)*DT*ACCELEIN); %Velocity
increment
VELOCOUT=VELOCIN+VARVEL; %Velocity of time step i+1
%
% Non-linear or linear stiffness force, internal spring force and state variable of time step
i+1
%[KEYSTIOUT,SPRFOROUT,fs(:,j+1),UTOUT,UCOUT,UOOUT,EMKEY]=f9_lin_stiff_int_force(NDOF,KEYSTIN,DISP
LAIN,u(:,j+1),SPRFORIN,KELAST);
%[KEYSTIOUT,SPRFOROUT,fs(:,j+1),UTOUT,UCOUT,UOOUT,EMKEY]=f10_nonlin_hyster_stiff_int_force_v3(VEL
OCOUT,NDOF,KEYSTIN,DISPLAIN,u(:,j+1),SPRFORIN,UTIN,UCIN,UOIN,ITIME,KELAST);
[KEYSTIOUT,SPRFOROUT,fs(:,j+1),UTOUT,UCOUT,UOOUT,EMKEY]=f11_nonlin_hyster_stiff_int_force_cubic(V
ELOCOUT,NDOF,KEYSTIN,DISPLAIN,u(:,j+1),UTIN,UCIN,UOIN,ITIME,KELAST);
if EMKEY~=100
return
end
%
%Nonlinear damping force and internal dashpot force of time step i+1
%[fd(:,j+1),DASHFOUT]=f12_lin_damp_int_force_v2(NDOF,VARVEL,DASHFOIN,CELAST);
[fd(:,j+1),DASHFOUT]=f13_nonlinmet_damp_int_force(NDOF,u(:,j+1),VELOCOUT,CELAST);
%
%Update the residual load increment for the next iteration
if j==1
dr(:,1)=VARLOD+(1/beta/DT)*M*VELOCIN+(1/2/beta)*M*ACCELEIN;
end
dr(:,j+1)=dr(:,j)-(fs(:,j+1)-fs(:,j)+fd(:,j+1)-fd(:,j)+(1/beta/(DT^2))*M*du(:,j));
%
%Error calculation
error=norm(du(:,j))/norm(VARDIS);
end
%
%Calculation of the system response for the time step i+1
DISPLAOUT=u(:,j+1);
FNLSTIOUT=fs(:,j+1);
FNLDAMOUT=fd(:,j+1);

216

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

ACCELEOUT=M\(LODFAC*P-FNLDAMOUT-FNLSTIOUT);
NITERA=j;

Published with MATLAB R2012b

217

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

B.3. Internal force models

B.3.1. Stiffness force functions

function
[KEYOUT,INTFOROUT,FNLSTIOUT,UTOUT,UCOUT,UOOUT,EMKEY]=f9_lin_stiff_int_force(NDOF,KEYIN,DISPLA1,DI
SPLA2,INTFORIN,KELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Stiffness force models
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal stiffness force corresponding to
%
the linear elastic model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system.
%
KEYIN
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
DISPLA1
Displacement vector of the previous time step
%
DISPLA2
Displacement vector of the current time step
%
INTFORIN
Spring internal stiffness force of the previous time
%
step
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
%
%
outputs
%
%
KEYOUT
Vector speciying the nonlinear state of each degree of
%
freedom in the current time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
INTFOROUT
Spring internal stiffness force of the current time
%
step
%
FNLSIOUT
Non-linear restoring stiffness force vector of the
%
current time step
%
%

218

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Calculation of relative displacements increments
DISREL1=zeros(NDOF+1,1);
DISREL2=zeros(NDOF+1,1);
DISREL1(1)=DISPLA1(1);
DISREL1(NDOF+1)=-DISPLA1(NDOF);
DISREL2(1)=DISPLA2(1);
DISREL2(NDOF+1)=-DISPLA2(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL1(IDOF)=DISPLA1(IDOF)-DISPLA1(IDOF-1);
DISREL2(IDOF)=DISPLA2(IDOF)-DISPLA2(IDOF-1);
end
end
VARDISREL=DISREL2-DISREL1;
%
% Defining the nonlinear state and non-linear internal spring force for
% each degree of freedom in the current time step
EMKEY=100;
KEYOUT=zeros(NDOF+1,1);
INTFOROUT=zeros(NDOF+1,1);
UTOUT=zeros(NDOF+1,1);
UCOUT=zeros(NDOF+1,1);
UOOUT=zeros(NDOF+1,1);
for IDOF=1:NDOF+1
if KELAST(IDOF)==0
INTFOROUT(IDOF)=0;
elseif KEYIN(IDOF)==0
INTFOROUT(IDOF)=INTFORIN(IDOF)+KELAST(IDOF)*VARDISREL(IDOF);
end
end
%
% Recalculation of the nonlinear restoring force vector for the current
% time step
FNLSTIOUT=zeros(NDOF,1);
for IDOF=1:NDOF
FNLSTIOUT(IDOF)=INTFOROUT(IDOF)-INTFOROUT(IDOF+1);
end

Published with MATLAB R2012b

219

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

function
[KEYOUT,INTFOROUT,FNLSTIOUT,UTOUT,UCOUT,UOOUT,EMKEY]=f10_nonlin_hyster_stiff_int_force_v3(VELOCI,
NDOF,KEYIN,DISPLA1,DISPLA2,INTFORIN,UTIN,UCIN,UOIN,ITIME,KELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Stiffness force models
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal stiffness force corresponding to
%
the elastoplastic hysteretic model
%
%
inputs
%
%
%
VELOCI
Velocity vector of the current time step.
%
NDOF
Number of degrees of freedom of the system.
%
KEYIN
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
DISPLA1
Displacement vector of the previous time step
%
DISPLA2
Displacement vector of the current time step
%
INTFORIN
Spring internal stiffness force of the previous time
%
step
%
UTIN
Maximum tensile displacement in the elastic range of
%
the previous time step
%
UCIN
Maximum compression displacement in the elastic range
%
of the previous time step
%
UOIN
Zero stress displacement of the previous time step
%
ITIME
Current time step index
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
%
%
outputs
%
%
KEYOUT
Vector speciying the nonlinear state of each degree of
%
freedom in the current time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
INTFOROUT
Spring internal stiffness force of the current time
%
step
%
FNLSIOUT
Non-linear restoring stiffness force vector of the
%
current time step
%
UTOUT
Maximum tensile displacement in the elastic range of
%
the current time step

220

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%
UCOUT
Maximum compression displacement in the elastic range
%
of the current time step
%
UOOUT
Zero stress displacement of the current time step
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Define the nonlinear elasto-plastic parameters for each degree of
% freedom
% This data has to be modified by the user depending of the problem data
%
TENSTR=[7.5 7.5 12.35 12.35 0 0 0 0 0 0];% Tensile strength for each degree of freedom
COMSTR=[-7.5 -7.5 12.35 12.35 0 0 0 0 0 0];% Compressive strength for each degree of freedom
if ITIME==1
for IDOF=1:NDOF+1
UTIN(IDOF)=TENSTR(IDOF)/KELAST(IDOF);
UCIN(IDOF)=COMSTR(IDOF)/KELAST(IDOF);
end
sizet=size(TENSTR);
sizec=size(COMSTR);
sizeut=size(UTIN);
sizeuc=size(UCIN);
sizeuo=size(UOIN);
%
% Checking if the number of parameters provided are enough
%
if sizet(2)<NDOF+1 || sizec(2)<NDOF+1
disp('Error message: the number coefficients of the vectors "TENSTR" and COMSTR, defined
in function "nonlin_hyster_stiff_int_force...", should be equal or higher to NDOF +1')
elseif sizeut(2)<NDOF+1 || sizeuc(2)<NDOF+1 || sizeuo(2)<NDOF+1
disp('Error message: the number coefficients of the vectors "UTIN", "UCIN" AND "UOIN",
defined in function "nonlin_hyster_stiff_int_force", should be equal or higher to NDOF +1')
end
end
%
% Calculation of relative displacements increments and relative velocities
%
DISREL1=zeros(NDOF+1,1);
DISREL2=zeros(NDOF+1,1);
VELREL=zeros(NDOF+1,1);
DISREL1(1)=DISPLA1(1);
DISREL1(NDOF+1)=-DISPLA1(NDOF);
DISREL2(1)=DISPLA2(1);
DISREL2(NDOF+1)=-DISPLA2(NDOF);
VELREL(1)=VELOCI(1);
VELREL(NDOF+1)=-VELOCI(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL1(IDOF)=DISPLA1(IDOF)-DISPLA1(IDOF-1);
DISREL2(IDOF)=DISPLA2(IDOF)-DISPLA2(IDOF-1);

221

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

VELREL(IDOF)=VELOCI(IDOF)-VELOCI(IDOF-1);
end
end
VARDISREL=DISREL2-DISREL1;
%
% Defining the nonlinear state and nonlinear internal force for each degree
% of freedom in the current time step
EMKEY=100;
KEYOUT=zeros(NDOF+1,1);
INTFOROUT=zeros(NDOF+1,1);
UTOUT=UTIN;
UCOUT=UCIN;
UOOUT=UOIN;
for IDOF=1:NDOF+1
if KELAST(IDOF)==0
INTFOROUT(IDOF)=0;
elseif KEYIN(IDOF)==0
INTFOROUT(IDOF)=INTFORIN(IDOF)+KELAST(IDOF)*VARDISREL(IDOF);
if INTFOROUT(IDOF)>=TENSTR(IDOF)
KEYOUT(IDOF)=1;
INTFOROUT(IDOF)=TENSTR(IDOF);
elseif INTFOROUT(IDOF)<=COMSTR(IDOF)
KEYOUT(IDOF)=-1;
INTFOROUT(IDOF)=COMSTR(IDOF);
else
KEYOUT(IDOF)=0;
end
elseif KEYIN(IDOF)==1
INTFOROUT(IDOF)=TENSTR(IDOF);
if VELREL(IDOF)>0
KEYOUT(IDOF)=1;
else
KEYOUT(IDOF)=0;
if VARDISREL(IDOF)<0
INTFOROUT(IDOF)=TENSTR(IDOF)+KELAST(IDOF)*VARDISREL(IDOF);
UTOUT(IDOF)=DISREL1(IDOF);
else
UTOUT(IDOF)=DISREL2(IDOF);
end
UCOUT(IDOF)=UTOUT(IDOF)-(TENSTR(IDOF)-COMSTR(IDOF))/KELAST(IDOF);
UOOUT(IDOF)=UTOUT(IDOF)-TENSTR(IDOF)/KELAST(IDOF);
end
else
INTFOROUT(IDOF)=COMSTR(IDOF);
if VELREL(IDOF)<0
KEYOUT(IDOF)=-1;
else
KEYOUT(IDOF)=0;
if VARDISREL(IDOF)>0
INTFOROUT(IDOF)=COMSTR(IDOF)+KELAST(IDOF)*VARDISREL(IDOF);

222

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

UCOUT(IDOF)=DISREL1(IDOF);
else
UCOUT(IDOF)=DISREL2(IDOF);
end
UTOUT(IDOF)=UCOUT(IDOF)+(TENSTR(IDOF)-COMSTR(IDOF))/KELAST(IDOF);
UOOUT(IDOF)=UTOUT(IDOF)-TENSTR(IDOF)/KELAST(IDOF);
end
end
end
%
% Recalculation of the nonlinear restoring force vector for the current time
% step
FNLSTIOUT=zeros(NDOF,1);
for IDOF=1:NDOF
FNLSTIOUT(IDOF)=INTFOROUT(IDOF)-INTFOROUT(IDOF+1);
end

Published with MATLAB R2012b

223

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

function
[KEYOUT,INTFOROUT,FNLSTIOUT,UTOUT,UCOUT,UOOUT,EMKEY]=f11_nonlin_hyster_stiff_int_force_cubic(VELO
CI,NDOF,KEYIN,DISPLA1,DISPLA2,UTIN,UCIN,UOIN,ITIME,KELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Stiffness force models
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal stiffness force corresponding to
%
the cubic elastoplastic hysteretic model
%
%
%
inputs
%
%
VELOCI
Velocity vector of the current time step.
%
NDOF
Number of degrees of freedom of the system.
%
KEYIN
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
DISPLA1
Displacement vector of the previous time step
%
DISPLA2
Displacement vector of the current time step
%
UTIN
Maximum tensile displacement in the elastic range of
%
the previous time step
%
UCIN
Maximum compression displacement in the elastic range
%
of the previous time step
%
UOIN
Zero stress displacement of the previous time step
%
ITIME
Current time step index
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
%
%
outputs
%
%
KEYOUT
Vector speciying the nonlinear state of each degree of
%
freedom in the current time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=-1 Plastic behavior in compression
%
INTFOROUT
Spring internal stiffness force of the current time
%
step
%
FNLSIOUT
Non-linear restoring stiffness force vector of the
%
current time step
%
UTOUT
Maximum tensile displacement in the elastic range of
%
the current time step
%
UCOUT
Maximum compression displacement in the elastic range
%
of the current time step

224

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

%
UOOUT
Zero stress displacement of the current time step
%
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Define the nonlinear cubic elasto-plastic parameters for each degree of
% freedom
% Cubic non-linearity
fs=k1*u-k3*u^3
% This data has to be modified by the user depending of the problem data
%
k1=[8 8 8 8 8 8 8 8 8 8];
k3=[32/27 32/27 32/27 32/27 32/27 32/27 32/27 32/27 32/27 32/27];
d=sqrt(k1./(3*k3));
TENSTR=(k1.*d)-k3.*(d.^3);% Tensile strength for each degree of freedom
COMSTR=-TENSTR;% Compressive strength for each degree of freedom
if ITIME==1
UTIN=transp(d);
UCIN=-UTIN;
sizet=size(TENSTR);
sizec=size(COMSTR);
sizeut=size(UTIN);
sizeuc=size(UCIN);
sizeuo=size(UOIN);
%
% Checking if the number of parameters provided are enough
%
if sizet(2)<NDOF+1 || sizec(2)<NDOF+1
disp('Error message: the number coefficients of the vectors "TENSTR" and COMSTR, defined
in function "nonlin_hyster_stiff_int_force...", should be equal or higher to NDOF +1')
elseif sizeut(2)<NDOF+1 || sizeuc(2)<NDOF+1 || sizeuo(2)<NDOF+1
disp('Error message: the number coefficients of the vectors "UTIN", "UCIN" AND "UOIN",
defined in function "nonlin_hyster_stiff_int_force", should be equal or higher to NDOF +1')
end
end
%
% Calculation of relative displacements increments and relative velocities
DISREL1=zeros(NDOF+1,1);
DISREL2=zeros(NDOF+1,1);
VELREL=zeros(NDOF+1,1);
DISREL1(1)=DISPLA1(1);
DISREL1(NDOF+1)=-DISPLA1(NDOF);
DISREL2(1)=DISPLA2(1);
DISREL2(NDOF+1)=-DISPLA2(NDOF);
VELREL(1)=VELOCI(1);
VELREL(NDOF+1)=-VELOCI(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL1(IDOF)=DISPLA1(IDOF)-DISPLA1(IDOF-1);
DISREL2(IDOF)=DISPLA2(IDOF)-DISPLA2(IDOF-1);

225

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

VELREL(IDOF)=VELOCI(IDOF)-VELOCI(IDOF-1);
end
end
VARDISREL=DISREL2-DISREL1;
%
% Defining the nonlinear state and nonlinear internal force for each degree
% of freedom in the current time step
EMKEY=100; %Value of state variable for error message
KEYOUT=zeros(NDOF+1,1);
INTFOROUT=zeros(NDOF+1,1);
UTOUT=UTIN;
UCOUT=UCIN;
UOOUT=UOIN;
for IDOF=1:NDOF+1
if KELAST(IDOF)==0
INTFOROUT(IDOF)=0;
elseif KEYIN(IDOF)==0
if DISREL2(IDOF)>=UTIN(IDOF)
KEYOUT(IDOF)=1;
INTFOROUT(IDOF)=TENSTR(IDOF);
elseif DISREL2(IDOF)<=UCIN(IDOF)
KEYOUT(IDOF)=-1;
INTFOROUT(IDOF)=COMSTR(IDOF);
else
KEYOUT(IDOF)=0;
INTFOROUT(IDOF)=k1(IDOF)*(DISREL2(IDOF)-UOIN(IDOF))-k3(IDOF)*((DISREL2(IDOF)UOIN(IDOF))^3);
end
elseif KEYIN(IDOF)==1
INTFOROUT(IDOF)=TENSTR(IDOF);
if VELREL(IDOF)>0
KEYOUT(IDOF)=1;
else
KEYOUT(IDOF)=2;
if VARDISREL(IDOF)<0
INTFOROUT(IDOF)=TENSTR(IDOF)+KELAST(IDOF)*VARDISREL(IDOF);
UTOUT(IDOF)=DISREL1(IDOF);
else
UTOUT(IDOF)=DISREL2(IDOF);
end
UOOUT(IDOF)=UTOUT(IDOF)-TENSTR(IDOF)/KELAST(IDOF);
UCOUT(IDOF)=UOOUT(IDOF)-d(IDOF);
end
elseif KEYIN(IDOF)==2
if DISREL2(IDOF)<=UOIN(IDOF)
KEYOUT(IDOF)=3;
INTFOROUT(IDOF)=k1(IDOF)*(DISREL2(IDOF)-UOIN(IDOF))-k3(IDOF)*((DISREL2(IDOF)UOIN(IDOF))^3);
elseif DISREL2(IDOF)>= UTIN(IDOF)
KEYOUT(IDOF)=1;

226

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

INTFOROUT(IDOF)=TENSTR(IDOF);
else
KEYOUT(IDOF)=2;
INTFOROUT(IDOF)=KELAST(IDOF)*(DISREL2(IDOF)-UOIN(IDOF));
end
elseif KEYIN(IDOF)==3
if DISREL2(IDOF)<=UCIN(IDOF)
KEYOUT(IDOF)=-1;
INTFOROUT(IDOF)=COMSTR(IDOF);
elseif DISREL2(IDOF)>= UOIN(IDOF)
KEYOUT(IDOF)=2;
INTFOROUT(IDOF)=KELAST(IDOF)*(DISREL2(IDOF)-UOIN(IDOF));
else
KEYOUT(IDOF)=3;
INTFOROUT(IDOF)=k1(IDOF)*(DISREL2(IDOF)-UOIN(IDOF))-k3(IDOF)*((DISREL2(IDOF)UOIN(IDOF))^3);
end
else
EMKEY=KEYIN(IDOF);
SPRINGNUM=IDOF;
fprintf('\nNot defined state variable KEYIN=%d of spring number %d for the calculation of
time step %d\n',EMKEY,SPRINGNUM,ITIME+1)
return
end
end
%
% Recalculation of the nonlinear restoring force vector for the current time
% step
FNLSTIOUT=zeros(NDOF,1);
for IDOF=1:NDOF
FNLSTIOUT(IDOF)=INTFOROUT(IDOF)-INTFOROUT(IDOF+1);
end

Published with MATLAB R2012b

227

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

B.3.2. Damping force functions

function [FNLDAMOUT,INTFOROUT]=f12_lin_damp_int_force_v2(NDOF,VARVEL,INTFORIN,CELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Damping force models
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal damping force corresponding to
%
the linear model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system
%
VARVEL
Velocity increment vector from the previous to the
%
current time step
%
INTFORIN
Dashpot internal damping force of the previous time
%
step
%
CELAST
Vector containing the linear damping coefficients of
%
the dashpots
%
%
%
outputs
%
%
INTFOROUT
Dashpot internal damping force of the current time
%
step
%
FNLDAMOUT
Non-linear restoring damping force vector of the
%
current time step
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Calculating the relative variation of velocity
VARVELREL=zeros(NDOF+1,1);
VARVELREL(1)=VARVEL(1);
VARVELREL(NDOF+1)=-VARVEL(NDOF);
if NDOF>=2
for IDOF=2:NDOF
VARVELREL(IDOF)=VARVEL(IDOF)-VARVEL(IDOF-1);
end
end
%
% Calculation of dashpot force
INTFOROUT=zeros(NDOF+1,1);
for IDOF=1:NDOF+1
INTFOROUT(IDOF)=INTFORIN(IDOF)+CELAST(IDOF)*VARVELREL(IDOF);

228

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

end
%
% Calculation of nonlinear internal damping force
FNLDAMOUT=zeros(NDOF,1);
for IDOF=1:NDOF
FNLDAMOUT(IDOF)=INTFOROUT(IDOF)-INTFOROUT(IDOF+1);
end

Published with MATLAB R2012b

229

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

function [FNLDAMOUT,INTFOROUT]=f13_nonlinmet_damp_int_force(NDOF,DISPLA,VELOCI,CELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Damping force models
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal damping force corresponding to
%
the non-linear displacement dependent model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system
%
DISPLA
Displacement vector of the current time step
%
VELOCI
Velocity vector of the current time step
%
CELAST
Vector containing the linear damping coefficients of
%
the dashpots
%
%
%
outputs
%
%
INTFOROUT
Dashpot internal damping force of the current time
%
step
%
FNLDAMOUT
Non-linear restoring damping force vector of the
%
current time step
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Quadratic displacement dependent non-linearity
%
fd=CELAST*VELOCI(ALFA*DISPLA^2+1)
%
%Nonlinear damping parameters
alfa=1;
%
%Relative displacement
DISREL=zeros(NDOF+1,1);
DISREL(1)=DISPLA(1);
DISREL(NDOF+1)=-DISPLA(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL(IDOF)=DISPLA(IDOF)-DISPLA(IDOF-1);
end
end
%
%Relative velocity
VELREL=zeros(NDOF+1,1);
VELREL(1)=VELOCI(1);
VELREL(NDOF+1)=-VELOCI(NDOF);

230

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

if NDOF>=2
for IDOF=2:NDOF
VELREL(IDOF)=VELOCI(IDOF)-VELOCI(IDOF-1);
end
end
%
% Calculation of dashpot force
INTFOROUT=zeros(NDOF+1,1);
for IDOF=1:NDOF+1
INTFOROUT(IDOF)=VELREL(IDOF)*CELAST(IDOF)*(alfa*((DISREL(IDOF))^2)+1);
%INTFOROUT(IDOF)=VELREL(IDOF)*CELAST(IDOF)*(alfa*((DISREL(IDOF))^2)-1);
end
%
% Calculation of nonlinear internal damping force
FNLDAMOUT=zeros(NDOF,1);
for IDOF=1:NDOF
FNLDAMOUT(IDOF)=INTFOROUT(IDOF)-INTFOROUT(IDOF+1);
end

Published with MATLAB R2012b

231

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix B: MATLAB Routines

232

Numerical modeling of dam-reservoir interaction seismic response


using the Hybrid FrequencyTime Domain (HFTD) method

Appendix C: DIANA HFTD Command File

APPENDIX C: DIANA HFTD COMMAND FILE

*FILOS
INITIA
*INPUT
*PHASE
BEGIN ACTIVE
ELEMEN "Dam" "Found" /
REINFO
END ACTIVE
*NONLIN
BEGIN EXECUT
BEGIN LOAD
LOADNR 1
STEPS EXPLIC SIZES 0.25(4)
END LOAD
ITERAT CONVER SIMULT
END EXECUT
*PHASE
*HFTD
BEGIN EIGEN
EXECUT NMODES 15
BEGIN OUTPUT
FXPLUS
FILE "eigenHFTD_dry"
END OUTPUT
END EIGEN
BEGIN MODEL
BEGIN MATRIX
DAMPIN
STRESS PHASE
END MATRIX
END MODEL
BEGIN RESPON
BEGIN EXECUT
ITERAT MAXITE 50
MAXFRE 25
TIMESE STEPS EXPLIC NUMBER 50
END EXECUT
BEGIN OUTPUT
FXPLUS
FILE "respHFTD_wet"
ACCELE TOTAL TRANSL GLOBAL
DISPLA TOTAL TRANSL GLOBAL
FORCE REACTI TRANSL GLOBAL
FSPRES TOTAL
STATUS CRACK
STRAIN TOTAL GREEN GLOBAL
STRAIN CRACK GREEN
STRESS TOTAL CAUCHY GLOBAL
VELOCI TOTAL TRANSL GLOBAL
END OUTPUT
REDUCE SELECT MODES 1-15 /
END RESPON
*END
233